lotus/node/impl/client/client.go

1536 lines
44 KiB
Go
Raw Normal View History

2019-09-16 13:46:05 +00:00
package client
2019-08-20 16:48:33 +00:00
import (
"bufio"
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
"bytes"
2019-08-20 16:48:33 +00:00
"context"
2021-11-23 16:42:43 +00:00
"errors"
"fmt"
2019-10-23 09:18:22 +00:00
"io"
2019-08-20 16:48:33 +00:00
"os"
"sort"
"strings"
"sync"
2021-05-11 02:26:04 +00:00
"time"
2019-08-20 16:48:33 +00:00
2022-06-14 15:00:51 +00:00
"github.com/ipfs/go-blockservice"
"github.com/ipfs/go-cid"
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
bstore "github.com/ipfs/go-ipfs-blockstore"
2022-06-14 15:00:51 +00:00
offline "github.com/ipfs/go-ipfs-exchange-offline"
files "github.com/ipfs/go-ipfs-files"
2021-11-11 15:17:39 +00:00
format "github.com/ipfs/go-ipld-format"
2022-06-14 15:00:51 +00:00
logging "github.com/ipfs/go-log/v2"
"github.com/ipfs/go-merkledag"
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
unixfile "github.com/ipfs/go-unixfs/file"
"github.com/ipld/go-car"
"github.com/ipld/go-car/util"
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
carv2 "github.com/ipld/go-car/v2"
carv2bs "github.com/ipld/go-car/v2/blockstore"
"github.com/ipld/go-ipld-prime"
2022-06-14 15:00:51 +00:00
"github.com/ipld/go-ipld-prime/datamodel"
cidlink "github.com/ipld/go-ipld-prime/linking/cid"
2020-07-07 08:52:19 +00:00
basicnode "github.com/ipld/go-ipld-prime/node/basic"
"github.com/ipld/go-ipld-prime/traversal"
2020-07-07 08:52:19 +00:00
"github.com/ipld/go-ipld-prime/traversal/selector"
"github.com/ipld/go-ipld-prime/traversal/selector/builder"
2021-10-07 08:47:51 +00:00
selectorparse "github.com/ipld/go-ipld-prime/traversal/selector/parse"
textselector "github.com/ipld/go-ipld-selector-text-lite"
2022-08-25 18:20:41 +00:00
"github.com/libp2p/go-libp2p/core/host"
"github.com/libp2p/go-libp2p/core/peer"
"github.com/multiformats/go-multibase"
2019-08-20 16:48:33 +00:00
"go.uber.org/fx"
2022-06-14 15:00:51 +00:00
"golang.org/x/xerrors"
2019-09-06 22:39:47 +00:00
"github.com/filecoin-project/go-address"
cborutil "github.com/filecoin-project/go-cbor-util"
"github.com/filecoin-project/go-commp-utils/writer"
datatransfer "github.com/filecoin-project/go-data-transfer"
2020-09-29 11:53:30 +00:00
"github.com/filecoin-project/go-fil-markets/discovery"
2020-06-23 19:22:33 +00:00
rm "github.com/filecoin-project/go-fil-markets/retrievalmarket"
"github.com/filecoin-project/go-fil-markets/storagemarket"
"github.com/filecoin-project/go-fil-markets/storagemarket/network"
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
"github.com/filecoin-project/go-fil-markets/stores"
2022-06-14 15:00:51 +00:00
"github.com/filecoin-project/go-padreader"
2020-09-07 03:49:10 +00:00
"github.com/filecoin-project/go-state-types/abi"
2022-06-14 15:00:51 +00:00
"github.com/filecoin-project/go-state-types/big"
2022-09-26 16:47:48 +00:00
markettypes "github.com/filecoin-project/go-state-types/builtin/v9/market"
2022-06-14 15:00:51 +00:00
"github.com/filecoin-project/go-state-types/dline"
2020-08-11 20:04:00 +00:00
"github.com/filecoin-project/lotus/api"
"github.com/filecoin-project/lotus/build"
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
"github.com/filecoin-project/lotus/chain/actors/builtin/miner"
"github.com/filecoin-project/lotus/chain/store"
"github.com/filecoin-project/lotus/chain/types"
2022-06-14 15:00:51 +00:00
"github.com/filecoin-project/lotus/lib/unixfs"
"github.com/filecoin-project/lotus/markets/retrievaladapter"
"github.com/filecoin-project/lotus/markets/storageadapter"
"github.com/filecoin-project/lotus/markets/utils"
2022-06-14 15:00:51 +00:00
"github.com/filecoin-project/lotus/node/config"
"github.com/filecoin-project/lotus/node/impl/full"
"github.com/filecoin-project/lotus/node/impl/paych"
2020-07-07 08:52:19 +00:00
"github.com/filecoin-project/lotus/node/modules/dtypes"
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
"github.com/filecoin-project/lotus/node/repo"
2022-06-14 15:00:51 +00:00
"github.com/filecoin-project/lotus/node/repo/imports"
2019-08-20 16:48:33 +00:00
)
var log = logging.Logger("client")
2022-03-21 09:37:35 +00:00
var DefaultHashFunction = unixfs.DefaultHashFunction
2020-07-07 09:38:22 +00:00
// 8 days ~= SealDuration + PreCommit + MaxProveCommitDuration + 8 hour buffer
const dealStartBufferHours uint64 = 8 * 24
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
const DefaultDAGStoreDir = "dagstore"
2019-09-16 13:46:05 +00:00
type API struct {
2019-08-20 16:48:33 +00:00
fx.In
2019-09-16 13:46:05 +00:00
full.ChainAPI
full.WalletAPI
paych.PaychAPI
full.StateAPI
2019-08-20 16:48:33 +00:00
SMDealClient storagemarket.StorageClient
2020-09-29 11:53:30 +00:00
RetDiscovery discovery.PeerResolver
2020-06-23 19:22:33 +00:00
Retrieval rm.RetrievalClient
2019-09-06 22:39:47 +00:00
Chain *store.ChainStore
2019-08-20 16:48:33 +00:00
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
// accessors for imports and retrievals.
Imports dtypes.ClientImportMgr
StorageBlockstoreAccessor storagemarket.BlockstoreAccessor
RtvlBlockstoreAccessor rm.BlockstoreAccessor
ApiBlockstoreAccessor *retrievaladapter.APIBlockstoreAccessor
2020-07-07 12:35:02 +00:00
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
DataTransfer dtypes.ClientDataTransfer
Host host.Host
Repo repo.LockedRepo
2019-08-20 16:48:33 +00:00
}
2020-09-10 06:30:47 +00:00
func calcDealExpiration(minDuration uint64, md *dline.Info, startEpoch abi.ChainEpoch) abi.ChainEpoch {
2020-04-21 21:38:26 +00:00
// Make sure we give some time for the miner to seal
2020-04-30 17:42:16 +00:00
minExp := startEpoch + abi.ChainEpoch(minDuration)
2020-04-21 21:38:26 +00:00
// Align on miners ProvingPeriodBoundary
exp := minExp + md.WPoStProvingPeriod - (minExp % md.WPoStProvingPeriod) + (md.PeriodStart % md.WPoStProvingPeriod) - 1
// Should only be possible for miners created around genesis
for exp < minExp {
exp += md.WPoStProvingPeriod
}
return exp
2020-04-21 21:38:26 +00:00
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
// importManager converts the injected type to the required type.
func (a *API) importManager() *imports.Manager {
2020-07-07 08:52:19 +00:00
return a.Imports
}
func (a *API) ClientStartDeal(ctx context.Context, params *api.StartDealParams) (*cid.Cid, error) {
return a.dealStarter(ctx, params, false)
}
func (a *API) ClientStatelessDeal(ctx context.Context, params *api.StartDealParams) (*cid.Cid, error) {
return a.dealStarter(ctx, params, true)
}
func (a *API) dealStarter(ctx context.Context, params *api.StartDealParams, isStateless bool) (*cid.Cid, error) {
if isStateless {
if params.Data.TransferType != storagemarket.TTManual {
return nil, xerrors.Errorf("invalid transfer type %s for stateless storage deal", params.Data.TransferType)
}
if !params.EpochPrice.IsZero() {
return nil, xerrors.New("stateless storage deals can only be initiated with storage price of 0")
}
} else if params.Data.TransferType == storagemarket.TTGraphsync {
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
bs, onDone, err := a.dealBlockstore(params.Data.Root)
if err != nil {
return nil, xerrors.Errorf("failed to find blockstore for root CID: %w", err)
}
2021-12-11 21:03:00 +00:00
if has, err := bs.Has(ctx, params.Data.Root); err != nil {
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
return nil, xerrors.Errorf("failed to query blockstore for root CID: %w", err)
} else if !has {
return nil, xerrors.Errorf("failed to find root CID in blockstore: %w", err)
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
onDone()
}
walletKey, err := a.StateAccountKey(ctx, params.Wallet, types.EmptyTSK)
if err != nil {
2021-06-25 08:48:47 +00:00
return nil, xerrors.Errorf("failed resolving params.Wallet addr (%s): %w", params.Wallet, err)
}
exist, err := a.WalletHas(ctx, walletKey)
2019-08-20 16:48:33 +00:00
if err != nil {
2021-06-25 08:48:47 +00:00
return nil, xerrors.Errorf("failed getting addr from wallet (%s): %w", params.Wallet, err)
}
if !exist {
return nil, xerrors.Errorf("provided address doesn't exist in wallet")
2019-08-20 16:48:33 +00:00
}
mi, err := a.StateMinerInfo(ctx, params.Miner, types.EmptyTSK)
2019-08-20 16:48:33 +00:00
if err != nil {
return nil, xerrors.Errorf("failed getting peer ID: %w", err)
}
md, err := a.StateMinerProvingDeadline(ctx, params.Miner, types.EmptyTSK)
if err != nil {
2020-07-07 22:29:45 +00:00
return nil, xerrors.Errorf("failed getting miner's deadline info: %w", err)
}
if uint64(params.Data.PieceSize.Padded()) > uint64(mi.SectorSize) {
return nil, xerrors.New("data doesn't fit in a sector")
}
2020-04-30 17:42:16 +00:00
dealStart := params.DealStartEpoch
if dealStart <= 0 { // unset, or explicitly 'epoch undefined'
ts, err := a.ChainHead(ctx)
if err != nil {
return nil, xerrors.Errorf("failed getting chain height: %w", err)
}
blocksPerHour := 60 * 60 / build.BlockDelaySecs
2020-10-02 17:31:21 +00:00
dealStart = ts.Height() + abi.ChainEpoch(dealStartBufferHours*blocksPerHour) // TODO: Get this from storage ask
}
networkVersion, err := a.StateNetworkVersion(ctx, types.EmptyTSK)
if err != nil {
return nil, xerrors.Errorf("failed to get network version: %w", err)
}
st, err := miner.PreferredSealProofTypeFromWindowPoStType(networkVersion, mi.WindowPoStProofType)
if err != nil {
return nil, xerrors.Errorf("failed to get seal proof type: %w", err)
}
// regular flow
if !isStateless {
providerInfo := utils.NewStorageProviderInfo(params.Miner, mi.Worker, mi.SectorSize, *mi.PeerId, mi.Multiaddrs)
result, err := a.SMDealClient.ProposeStorageDeal(ctx, storagemarket.ProposeStorageDealParams{
Addr: params.Wallet,
Info: &providerInfo,
Data: params.Data,
StartEpoch: dealStart,
EndEpoch: calcDealExpiration(params.MinBlocksDuration, md, dealStart),
Price: params.EpochPrice,
Collateral: params.ProviderCollateral,
Rt: st,
FastRetrieval: params.FastRetrieval,
VerifiedDeal: params.VerifiedDeal,
})
if err != nil {
return nil, xerrors.Errorf("failed to start deal: %w", err)
}
2019-08-20 16:48:33 +00:00
return &result.ProposalCid, nil
}
//
// stateless flow from here to the end
//
2022-09-26 16:47:48 +00:00
label, err := markettypes.NewLabelFromString(params.Data.Root.Encode(multibase.MustNewEncoder('u')))
2022-03-12 18:07:35 +00:00
if err != nil {
return nil, xerrors.Errorf("failed to encode label: %w", err)
}
2022-09-26 16:47:48 +00:00
dealProposal := &markettypes.DealProposal{
PieceCID: *params.Data.PieceCid,
PieceSize: params.Data.PieceSize.Padded(),
Client: walletKey,
Provider: params.Miner,
2022-03-12 18:07:35 +00:00
Label: label,
StartEpoch: dealStart,
EndEpoch: calcDealExpiration(params.MinBlocksDuration, md, dealStart),
StoragePricePerEpoch: big.Zero(),
ProviderCollateral: params.ProviderCollateral,
ClientCollateral: big.Zero(),
VerifiedDeal: params.VerifiedDeal,
}
if dealProposal.ProviderCollateral.IsZero() {
networkCollateral, err := a.StateDealProviderCollateralBounds(ctx, params.Data.PieceSize.Padded(), params.VerifiedDeal, types.EmptyTSK)
if err != nil {
return nil, xerrors.Errorf("failed to determine minimum provider collateral: %w", err)
}
dealProposal.ProviderCollateral = networkCollateral.Min
}
dealProposalSerialized, err := cborutil.Dump(dealProposal)
if err != nil {
return nil, xerrors.Errorf("failed to serialize deal proposal: %w", err)
}
dealProposalSig, err := a.WalletSign(ctx, walletKey, dealProposalSerialized)
if err != nil {
return nil, xerrors.Errorf("failed to sign proposal : %w", err)
}
2022-09-26 16:47:48 +00:00
dealProposalSigned := &markettypes.ClientDealProposal{
Proposal: *dealProposal,
ClientSignature: *dealProposalSig,
}
dStream, err := network.NewFromLibp2pHost(a.Host,
2021-05-11 02:26:04 +00:00
// params duplicated from .../node/modules/client.go
// https://github.com/filecoin-project/lotus/pull/5961#discussion_r629768011
network.RetryParameters(time.Second, 5*time.Minute, 15, 5),
).NewDealStream(ctx, *mi.PeerId)
if err != nil {
return nil, xerrors.Errorf("opening dealstream to %s/%s failed: %w", params.Miner, *mi.PeerId, err)
}
if err = dStream.WriteDealProposal(network.Proposal{
FastRetrieval: true,
DealProposal: dealProposalSigned,
Piece: &storagemarket.DataRef{
TransferType: storagemarket.TTManual,
Root: params.Data.Root,
PieceCid: params.Data.PieceCid,
PieceSize: params.Data.PieceSize,
},
}); err != nil {
return nil, xerrors.Errorf("sending deal proposal failed: %w", err)
}
resp, _, err := dStream.ReadDealResponse()
if err != nil {
return nil, xerrors.Errorf("reading proposal response failed: %w", err)
}
dealProposalIpld, err := cborutil.AsIpld(dealProposalSigned)
if err != nil {
return nil, xerrors.Errorf("serializing proposal node failed: %w", err)
}
if !dealProposalIpld.Cid().Equals(resp.Response.Proposal) {
return nil, xerrors.Errorf("provider returned proposal cid %s but we expected %s", resp.Response.Proposal, dealProposalIpld.Cid())
}
if resp.Response.State != storagemarket.StorageDealWaitingForData {
return nil, xerrors.Errorf("provider returned unexpected state %d for proposal %s, with message: %s", resp.Response.State, resp.Response.Proposal, resp.Response.Message)
}
return &resp.Response.Proposal, nil
2019-08-20 16:48:33 +00:00
}
2019-09-16 13:46:05 +00:00
func (a *API) ClientListDeals(ctx context.Context) ([]api.DealInfo, error) {
deals, err := a.SMDealClient.ListLocalDeals(ctx)
2019-09-10 14:13:24 +00:00
if err != nil {
return nil, err
}
// Get a map of transfer ID => DataTransfer
dataTransfersByID, err := a.transfersByID(ctx)
if err != nil {
return nil, err
}
2019-09-10 14:13:24 +00:00
out := make([]api.DealInfo, len(deals))
for k, v := range deals {
// Find the data transfer associated with this deal
var transferCh *api.DataTransferChannel
if v.TransferChannelID != nil {
if ch, ok := dataTransfersByID[*v.TransferChannelID]; ok {
transferCh = &ch
}
2019-09-10 14:13:24 +00:00
}
out[k] = a.newDealInfoWithTransfer(transferCh, v)
2019-09-10 14:13:24 +00:00
}
return out, nil
}
func (a *API) transfersByID(ctx context.Context) (map[datatransfer.ChannelID]api.DataTransferChannel, error) {
inProgressChannels, err := a.DataTransfer.InProgressChannels(ctx)
if err != nil {
return nil, err
}
dataTransfersByID := make(map[datatransfer.ChannelID]api.DataTransferChannel, len(inProgressChannels))
for id, channelState := range inProgressChannels {
ch := api.NewDataTransferChannel(a.Host.ID(), channelState)
dataTransfersByID[id] = ch
}
return dataTransfersByID, nil
}
2019-11-06 19:44:28 +00:00
func (a *API) ClientGetDealInfo(ctx context.Context, d cid.Cid) (*api.DealInfo, error) {
v, err := a.SMDealClient.GetLocalDeal(ctx, d)
2019-11-06 19:44:28 +00:00
if err != nil {
return nil, err
}
di := a.newDealInfo(ctx, v)
return &di, nil
2019-11-06 19:44:28 +00:00
}
2020-08-27 18:32:51 +00:00
func (a *API) ClientGetDealUpdates(ctx context.Context) (<-chan api.DealInfo, error) {
updates := make(chan api.DealInfo)
unsub := a.SMDealClient.SubscribeToEvents(func(_ storagemarket.ClientEvent, deal storagemarket.ClientDeal) {
updates <- a.newDealInfo(ctx, deal)
2020-08-27 18:32:51 +00:00
})
go func() {
defer unsub()
<-ctx.Done()
}()
return updates, nil
}
func (a *API) newDealInfo(ctx context.Context, v storagemarket.ClientDeal) api.DealInfo {
// Find the data transfer associated with this deal
var transferCh *api.DataTransferChannel
if v.TransferChannelID != nil {
state, err := a.DataTransfer.ChannelState(ctx, *v.TransferChannelID)
// Note: If there was an error just ignore it, as the data transfer may
// be not found if it's no longer active
if err == nil {
ch := api.NewDataTransferChannel(a.Host.ID(), state)
ch.Stages = state.Stages()
transferCh = &ch
}
}
di := a.newDealInfoWithTransfer(transferCh, v)
di.DealStages = v.DealStages
return di
}
func (a *API) newDealInfoWithTransfer(transferCh *api.DataTransferChannel, v storagemarket.ClientDeal) api.DealInfo {
return api.DealInfo{
ProposalCid: v.ProposalCid,
DataRef: v.DataRef,
State: v.State,
Message: v.Message,
Provider: v.Proposal.Provider,
PieceCID: v.Proposal.PieceCID,
Size: uint64(v.Proposal.PieceSize.Unpadded()),
PricePerEpoch: v.Proposal.StoragePricePerEpoch,
Duration: uint64(v.Proposal.Duration()),
DealID: v.DealID,
CreationTime: v.CreationTime.Time(),
Verified: v.Proposal.VerifiedDeal,
TransferChannelID: v.TransferChannelID,
DataTransfer: transferCh,
}
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
func (a *API) ClientHasLocal(_ context.Context, root cid.Cid) (bool, error) {
_, onDone, err := a.dealBlockstore(root)
2019-08-26 13:45:36 +00:00
if err != nil {
return false, err
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
onDone()
2019-08-26 13:45:36 +00:00
return true, nil
}
2020-07-09 16:29:57 +00:00
func (a *API) ClientFindData(ctx context.Context, root cid.Cid, piece *cid.Cid) ([]api.QueryOffer, error) {
2019-08-26 13:45:36 +00:00
peers, err := a.RetDiscovery.GetPeers(root)
if err != nil {
return nil, err
}
2020-07-09 16:29:57 +00:00
out := make([]api.QueryOffer, 0, len(peers))
for _, p := range peers {
if piece != nil && !piece.Equals(*p.PieceCID) {
continue
}
2021-07-21 09:43:19 +00:00
// do not rely on local data with respect to peer id
// fetch an up-to-date miner peer id from chain
mi, err := a.StateMinerInfo(ctx, p.Address, types.EmptyTSK)
if err != nil {
return nil, err
}
pp := rm.RetrievalPeer{
Address: p.Address,
ID: *mi.PeerId,
}
out = append(out, a.makeRetrievalQuery(ctx, pp, root, piece, rm.QueryParams{}))
2019-08-26 13:45:36 +00:00
}
return out, nil
}
func (a *API) ClientMinerQueryOffer(ctx context.Context, miner address.Address, root cid.Cid, piece *cid.Cid) (api.QueryOffer, error) {
mi, err := a.StateMinerInfo(ctx, miner, types.EmptyTSK)
if err != nil {
return api.QueryOffer{}, err
}
2020-06-23 19:22:33 +00:00
rp := rm.RetrievalPeer{
Address: miner,
ID: *mi.PeerId,
}
return a.makeRetrievalQuery(ctx, rp, root, piece, rm.QueryParams{}), nil
}
2020-07-09 16:29:57 +00:00
func (a *API) makeRetrievalQuery(ctx context.Context, rp rm.RetrievalPeer, payload cid.Cid, piece *cid.Cid, qp rm.QueryParams) api.QueryOffer {
queryResponse, err := a.Retrieval.Query(ctx, rp, payload, qp)
if err != nil {
2020-08-05 22:35:59 +00:00
return api.QueryOffer{Err: err.Error(), Miner: rp.Address, MinerPeer: rp}
}
var errStr string
switch queryResponse.Status {
2020-06-23 19:22:33 +00:00
case rm.QueryResponseAvailable:
errStr = ""
2020-06-23 19:22:33 +00:00
case rm.QueryResponseUnavailable:
errStr = fmt.Sprintf("retrieval query offer was unavailable: %s", queryResponse.Message)
2020-06-23 19:22:33 +00:00
case rm.QueryResponseError:
errStr = fmt.Sprintf("retrieval query offer errored: %s", queryResponse.Message)
}
return api.QueryOffer{
Root: payload,
2020-07-09 16:29:57 +00:00
Piece: piece,
Size: queryResponse.Size,
MinPrice: queryResponse.PieceRetrievalPrice(),
UnsealPrice: queryResponse.UnsealPrice,
2021-12-17 14:50:51 +00:00
PricePerByte: queryResponse.MinPricePerByte,
PaymentInterval: queryResponse.MaxPaymentInterval,
PaymentIntervalIncrease: queryResponse.MaxPaymentIntervalIncrease,
Miner: queryResponse.PaymentAddress, // TODO: check
2020-08-05 22:35:59 +00:00
MinerPeer: rp,
Err: errStr,
}
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
func (a *API) ClientImport(ctx context.Context, ref api.FileRef) (res *api.ImportRes, err error) {
var (
imgr = a.importManager()
id imports.ID
root cid.Cid
carPath string
)
id, err = imgr.CreateImport()
2020-07-06 23:39:30 +00:00
if err != nil {
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
return nil, xerrors.Errorf("failed to create import: %w", err)
2020-07-06 23:39:30 +00:00
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
if ref.IsCAR {
// user gave us a CAR file, use it as-is
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
// validate that it's either a carv1 or carv2, and has one root.
f, err := os.Open(ref.Path)
if err != nil {
return nil, xerrors.Errorf("failed to open CAR file: %w", err)
}
defer f.Close() //nolint:errcheck
2021-12-11 21:03:00 +00:00
hd, err := car.ReadHeader(bufio.NewReader(f))
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
if err != nil {
return nil, xerrors.Errorf("failed to read CAR header: %w", err)
}
if len(hd.Roots) != 1 {
return nil, xerrors.New("car file can have one and only one header")
}
if hd.Version != 1 && hd.Version != 2 {
return nil, xerrors.Errorf("car version must be 1 or 2, is %d", hd.Version)
}
carPath = ref.Path
root = hd.Roots[0]
} else {
carPath, err = imgr.AllocateCAR(id)
if err != nil {
return nil, xerrors.Errorf("failed to create car path for import: %w", err)
}
// remove the import if something went wrong.
defer func() {
if err != nil {
_ = os.Remove(carPath)
_ = imgr.Remove(id)
}
}()
// perform the unixfs chunking.
2022-03-21 09:37:35 +00:00
root, err = unixfs.CreateFilestore(ctx, ref.Path, carPath)
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
if err != nil {
return nil, xerrors.Errorf("failed to import file using unixfs: %w", err)
}
2020-07-06 23:39:30 +00:00
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
if err = imgr.AddLabel(id, imports.LSource, "import"); err != nil {
2020-07-07 11:45:02 +00:00
return nil, err
2020-07-07 09:38:09 +00:00
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
if err = imgr.AddLabel(id, imports.LFileName, ref.Path); err != nil {
2020-07-07 11:45:02 +00:00
return nil, err
2019-08-20 16:48:33 +00:00
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
if err = imgr.AddLabel(id, imports.LCARPath, carPath); err != nil {
return nil, err
}
if err = imgr.AddLabel(id, imports.LRootCid, root.String()); err != nil {
2020-07-07 11:45:02 +00:00
return nil, err
2020-07-07 09:38:09 +00:00
}
2020-07-07 11:45:02 +00:00
return &api.ImportRes{
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
Root: root,
2020-07-07 11:45:02 +00:00
ImportID: id,
}, nil
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
func (a *API) ClientRemoveImport(ctx context.Context, id imports.ID) error {
info, err := a.importManager().Info(id)
if err != nil {
return xerrors.Errorf("failed to get import metadata: %w", err)
}
owner := info.Labels[imports.LCAROwner]
path := info.Labels[imports.LCARPath]
// CARv2 file was not provided by the user, delete it.
if path != "" && owner == imports.CAROwnerImportMgr {
_ = os.Remove(path)
}
return a.importManager().Remove(id)
2019-08-20 16:48:33 +00:00
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
// ClientImportLocal imports a standard file into this node as a UnixFS payload,
// storing it in a CARv2 file. Note that this method is NOT integrated with the
// IPFS blockstore. That is, if client-side IPFS integration is enabled, this
// method won't import the file into that
func (a *API) ClientImportLocal(ctx context.Context, r io.Reader) (cid.Cid, error) {
file := files.NewReaderFile(r)
2019-10-23 09:18:22 +00:00
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
// write payload to temp file
id, err := a.importManager().CreateImport()
2020-07-06 23:39:30 +00:00
if err != nil {
2020-07-07 09:12:32 +00:00
return cid.Undef, err
2020-07-06 23:39:30 +00:00
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
if err := a.importManager().AddLabel(id, imports.LSource, "import-local"); err != nil {
return cid.Undef, err
}
path, err := a.importManager().AllocateCAR(id)
if err != nil {
return cid.Undef, err
2020-07-06 23:39:30 +00:00
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
// writing a carv2 requires knowing the root ahead of time, which makes
// streaming cases impossible.
// https://github.com/ipld/go-car/issues/196
// we work around this limitation by informing a placeholder root CID of the
// same length as our unixfs chunking strategy will generate.
// once the DAG is formed and the root is calculated, we overwrite the
// inner carv1 header with the final root.
2019-10-23 09:18:22 +00:00
2022-03-21 09:37:35 +00:00
b, err := unixfs.CidBuilder()
2020-07-07 09:12:32 +00:00
if err != nil {
return cid.Undef, err
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
// placeholder payload needs to be larger than inline CID threshold; 256
// bytes is a safe value.
placeholderRoot, err := b.Sum(make([]byte, 256))
if err != nil {
return cid.Undef, xerrors.Errorf("failed to calculate placeholder root: %w", err)
2019-10-23 09:18:22 +00:00
}
bs, err := carv2bs.OpenReadWrite(path, []cid.Cid{placeholderRoot}, carv2bs.UseWholeCIDs(true))
2019-10-23 09:18:22 +00:00
if err != nil {
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
return cid.Undef, xerrors.Errorf("failed to create carv2 read/write blockstore: %w", err)
2019-10-23 09:18:22 +00:00
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
2022-03-21 09:37:35 +00:00
root, err := unixfs.Build(ctx, file, bs, false)
2019-10-23 09:18:22 +00:00
if err != nil {
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
return cid.Undef, xerrors.Errorf("failed to build unixfs dag: %w", err)
}
err = bs.Finalize()
if err != nil {
return cid.Undef, xerrors.Errorf("failed to finalize carv2 read/write blockstore: %w", err)
}
// record the root in the import manager.
if err := a.importManager().AddLabel(id, imports.LRootCid, root.String()); err != nil {
return cid.Undef, xerrors.Errorf("failed to record root CID in import manager: %w", err)
}
// now go ahead and overwrite the root in the carv1 header.
reader, err := carv2.OpenReader(path)
if err != nil {
return cid.Undef, xerrors.Errorf("failed to create car reader: %w", err)
2019-10-23 09:18:22 +00:00
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
// save the header offset.
headerOff := reader.Header.DataOffset
// read the old header.
2022-07-06 10:11:30 +00:00
dr, err := reader.DataReader()
if err != nil {
return cid.Undef, fmt.Errorf("failed to get car data reader: %w", err)
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
header, err := readHeader(dr)
if err != nil {
return cid.Undef, xerrors.Errorf("failed to read car reader: %w", err)
}
_ = reader.Close() // close the CAR reader.
// write the old header into a buffer.
var oldBuf bytes.Buffer
if err = writeHeader(header, &oldBuf); err != nil {
return cid.Undef, xerrors.Errorf("failed to write header into buffer: %w", err)
}
// replace the root.
header.Roots = []cid.Cid{root}
// write the new header into a buffer.
var newBuf bytes.Buffer
err = writeHeader(header, &newBuf)
if err != nil {
return cid.Undef, xerrors.Errorf("failed to write header into buffer: %w", err)
}
// verify the length matches.
if newBuf.Len() != oldBuf.Len() {
return cid.Undef, xerrors.Errorf("failed to replace carv1 header; length mismatch (old: %d, new: %d)", oldBuf.Len(), newBuf.Len())
}
// open the file again, seek to the header position, and write.
f, err := os.OpenFile(path, os.O_WRONLY, 0755)
if err != nil {
return cid.Undef, xerrors.Errorf("failed to open car: %w", err)
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
defer f.Close() //nolint:errcheck
2019-10-23 09:18:22 +00:00
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
n, err := f.WriteAt(newBuf.Bytes(), int64(headerOff))
if err != nil {
return cid.Undef, xerrors.Errorf("failed to write new header to car (bytes written: %d): %w", n, err)
}
return root, nil
2019-10-23 09:18:22 +00:00
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
func (a *API) ClientListImports(_ context.Context) ([]api.Import, error) {
ids, err := a.importManager().List()
if err != nil {
return nil, xerrors.Errorf("failed to fetch imports: %w", err)
}
2019-08-20 16:48:33 +00:00
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
out := make([]api.Import, len(ids))
for i, id := range ids {
info, err := a.importManager().Info(id)
2020-07-07 08:52:19 +00:00
if err != nil {
out[i] = api.Import{
Key: id,
2020-07-07 09:12:32 +00:00
Err: xerrors.Errorf("getting info: %w", err).Error(),
2020-07-07 08:52:19 +00:00
}
continue
}
2019-08-20 16:48:33 +00:00
2020-07-07 08:52:19 +00:00
ai := api.Import{
Key: id,
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
Source: info.Labels[imports.LSource],
FilePath: info.Labels[imports.LFileName],
CARPath: info.Labels[imports.LCARPath],
2019-08-20 16:48:33 +00:00
}
2020-07-07 08:52:19 +00:00
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
if info.Labels[imports.LRootCid] != "" {
c, err := cid.Parse(info.Labels[imports.LRootCid])
2020-07-07 08:52:19 +00:00
if err != nil {
2020-07-07 09:12:32 +00:00
ai.Err = err.Error()
2020-07-07 08:52:19 +00:00
} else {
ai.Root = &c
}
}
2020-07-07 08:52:19 +00:00
out[i] = ai
2019-08-20 16:48:33 +00:00
}
2020-07-07 08:52:19 +00:00
return out, nil
2019-08-20 16:48:33 +00:00
}
2019-08-27 18:45:21 +00:00
func (a *API) ClientCancelRetrievalDeal(ctx context.Context, dealID rm.DealID) error {
2021-03-24 12:36:21 +00:00
cerr := make(chan error)
go func() {
err := a.Retrieval.CancelDeal(dealID)
2021-03-24 12:36:21 +00:00
select {
case cerr <- err:
case <-ctx.Done():
}
}()
select {
case err := <-cerr:
if err != nil {
return xerrors.Errorf("failed to cancel retrieval deal: %w", err)
2021-03-24 12:36:21 +00:00
}
return nil
case <-ctx.Done():
return xerrors.Errorf("context timeout while canceling retrieval deal: %w", ctx.Err())
2021-03-24 12:36:21 +00:00
}
}
func getDataSelector(dps *api.Selector, matchPath bool) (datamodel.Node, error) {
2021-10-07 08:47:51 +00:00
sel := selectorparse.CommonSelector_ExploreAllRecursively
2021-11-09 16:27:42 +00:00
if dps != nil {
2021-11-09 16:27:42 +00:00
if strings.HasPrefix(string(*dps), "{") {
var err error
2021-11-09 16:27:42 +00:00
sel, err = selectorparse.ParseJSONSelector(string(*dps))
if err != nil {
2021-11-09 16:27:42 +00:00
return nil, xerrors.Errorf("failed to parse json-selector '%s': %w", *dps, err)
}
} else {
ssb := builder.NewSelectorSpecBuilder(basicnode.Prototype.Any)
selspec, err := textselector.SelectorSpecFromPath(
textselector.Expression(*dps), matchPath,
ssb.ExploreRecursive(
selector.RecursionLimitNone(),
ssb.ExploreUnion(ssb.Matcher(), ssb.ExploreAll(ssb.ExploreRecursiveEdge())),
),
)
if err != nil {
2021-11-09 16:27:42 +00:00
return nil, xerrors.Errorf("failed to parse text-selector '%s': %w", *dps, err)
}
sel = selspec.Node()
2021-11-09 16:27:42 +00:00
log.Infof("partial retrieval of datamodel-path-selector %s/*", *dps)
}
}
2021-11-09 16:27:42 +00:00
return sel, nil
}
2019-09-16 20:11:17 +00:00
2021-11-10 14:45:46 +00:00
func (a *API) ClientRetrieve(ctx context.Context, params api.RetrievalOrder) (*api.RestrievalRes, error) {
sel, err := getDataSelector(params.DataSelector, false)
2021-11-10 14:45:46 +00:00
if err != nil {
return nil, err
}
di, err := a.doRetrieval(ctx, params, sel)
if err != nil {
return nil, err
}
return &api.RestrievalRes{
DealID: di,
}, nil
}
func (a *API) doRetrieval(ctx context.Context, order api.RetrievalOrder, sel datamodel.Node) (rm.DealID, error) {
2021-11-09 16:27:42 +00:00
if order.MinerPeer == nil || order.MinerPeer.ID == "" {
mi, err := a.StateMinerInfo(ctx, order.Miner, types.EmptyTSK)
if err != nil {
return 0, err
2019-09-16 20:11:17 +00:00
}
2021-11-09 16:27:42 +00:00
order.MinerPeer = &rm.RetrievalPeer{
ID: *mi.PeerId,
Address: order.Miner,
}
2021-11-09 16:27:42 +00:00
}
2021-04-23 08:10:51 +00:00
2021-11-09 16:27:42 +00:00
if order.Total.Int == nil {
return 0, xerrors.Errorf("cannot make retrieval deal for null total")
}
2019-09-16 20:11:17 +00:00
2021-11-09 16:27:42 +00:00
if order.Size == 0 {
return 0, xerrors.Errorf("cannot make retrieval deal for zero bytes")
}
2020-07-06 23:39:30 +00:00
ppb := types.BigDiv(big.Sub(order.Total, order.UnsealPrice), types.NewInt(order.Size))
2021-11-09 16:27:42 +00:00
params, err := rm.NewParamsV1(ppb, order.PaymentInterval, order.PaymentIntervalIncrease, sel, order.Piece, order.UnsealPrice)
if err != nil {
return 0, xerrors.Errorf("Error in retrieval params: %s", err)
}
id := a.Retrieval.NextID()
if order.RemoteStore != nil {
if err := a.ApiBlockstoreAccessor.UseRetrievalStore(id, *order.RemoteStore); err != nil {
return 0, xerrors.Errorf("registering api store: %w", err)
}
}
2021-11-09 16:27:42 +00:00
id, err = a.Retrieval.Retrieve(
ctx,
id,
order.Root,
params,
order.Total,
*order.MinerPeer,
order.Client,
order.Miner,
)
2021-11-09 16:27:42 +00:00
if err != nil {
return 0, xerrors.Errorf("Retrieve failed: %w", err)
2019-08-27 22:10:23 +00:00
}
2021-11-10 14:45:46 +00:00
return id, nil
}
2021-11-10 16:51:16 +00:00
func (a *API) ClientRetrieveWait(ctx context.Context, deal rm.DealID) error {
ctx, cancel := context.WithCancel(ctx)
defer cancel()
subscribeEvents := make(chan rm.ClientDealState, 1)
unsubscribe := a.Retrieval.SubscribeToEvents(func(event rm.ClientEvent, state rm.ClientDealState) {
// We'll check the deal IDs inside consumeAllEvents.
if state.ID != deal {
return
}
select {
case <-ctx.Done():
case subscribeEvents <- state:
}
})
defer unsubscribe()
{
state, err := a.Retrieval.GetDeal(deal)
if err != nil {
return xerrors.Errorf("getting deal state: %w", err)
}
select {
case subscribeEvents <- state:
default: // already have an event queued from the subscription
}
}
for {
select {
case <-ctx.Done():
return xerrors.New("Retrieval Timed Out")
case state := <-subscribeEvents:
switch state.Status {
case rm.DealStatusCompleted:
return nil
case rm.DealStatusRejected:
return xerrors.Errorf("Retrieval Proposal Rejected: %s", state.Message)
case rm.DealStatusCancelled:
return xerrors.Errorf("Retrieval was cancelled externally: %s", state.Message)
case
rm.DealStatusDealNotFound,
rm.DealStatusErrored:
return xerrors.Errorf("Retrieval Error: %s", state.Message)
}
}
}
}
2021-11-11 16:17:11 +00:00
type ExportDest struct {
Writer io.Writer
Path string
}
func (ed *ExportDest) doWrite(cb func(io.Writer) error) error {
if ed.Writer != nil {
return cb(ed.Writer)
}
f, err := os.OpenFile(ed.Path, os.O_CREATE|os.O_WRONLY, 0644)
if err != nil {
return err
}
if err := cb(f); err != nil {
_ = f.Close()
return err
}
return f.Close()
}
2021-11-10 14:45:46 +00:00
func (a *API) ClientExport(ctx context.Context, exportRef api.ExportRef, ref api.FileRef) error {
2021-11-11 16:17:11 +00:00
return a.ClientExportInto(ctx, exportRef, ref.IsCAR, ExportDest{Path: ref.Path})
}
func (a *API) ClientExportInto(ctx context.Context, exportRef api.ExportRef, car bool, dest ExportDest) error {
2021-11-10 14:45:46 +00:00
proxyBss, retrieveIntoIPFS := a.RtvlBlockstoreAccessor.(*retrievaladapter.ProxyBlockstoreAccessor)
carBss, retrieveIntoCAR := a.RtvlBlockstoreAccessor.(*retrievaladapter.CARBlockstoreAccessor)
carPath := exportRef.FromLocalCAR
2021-11-10 14:45:46 +00:00
if carPath == "" {
if !retrieveIntoIPFS && !retrieveIntoCAR {
return xerrors.Errorf("unsupported retrieval blockstore accessor")
}
if retrieveIntoCAR {
carPath = carBss.PathFor(exportRef.DealID)
}
}
var retrievalBs bstore.Blockstore
if retrieveIntoIPFS {
retrievalBs = proxyBss.Blockstore
} else {
cbs, err := stores.ReadOnlyFilestore(carPath)
if err != nil {
return err
}
defer cbs.Close() //nolint:errcheck
retrievalBs = cbs
}
2021-11-11 15:17:39 +00:00
dserv := merkledag.NewDAGService(blockservice.New(retrievalBs, offline.Exchange(retrievalBs)))
2021-11-10 14:45:46 +00:00
// Are we outputting a CAR?
2021-11-11 16:17:11 +00:00
if car {
2021-11-10 14:45:46 +00:00
// not IPFS and we do full selection - just extract the CARv1 from the CARv2 we stored the retrieval in
2021-11-11 16:17:11 +00:00
if !retrieveIntoIPFS && len(exportRef.DAGs) == 0 && dest.Writer == nil {
return carv2.ExtractV1File(carPath, dest.Path)
2021-11-10 14:45:46 +00:00
}
}
roots, err := parseDagSpec(ctx, exportRef.Root, exportRef.DAGs, dserv, car)
if err != nil {
return xerrors.Errorf("parsing dag spec: %w", err)
}
if car {
return a.outputCAR(ctx, dserv, retrievalBs, exportRef.Root, roots, dest)
2021-11-10 14:45:46 +00:00
}
2021-11-11 15:17:39 +00:00
if len(roots) != 1 {
return xerrors.Errorf("unixfs retrieval requires one root node, got %d", len(roots))
}
2021-11-11 16:17:11 +00:00
return a.outputUnixFS(ctx, roots[0].root, dserv, dest)
2021-11-09 16:27:42 +00:00
}
func (a *API) outputCAR(ctx context.Context, ds format.DAGService, bs bstore.Blockstore, root cid.Cid, dags []dagSpec, dest ExportDest) error {
2021-11-09 16:27:42 +00:00
// generating a CARv1 from the configured blockstore
roots := make([]cid.Cid, len(dags))
2021-11-11 15:17:39 +00:00
for i, dag := range dags {
roots[i] = dag.root
2021-11-11 15:17:39 +00:00
}
var lk sync.Mutex
2021-11-11 16:17:11 +00:00
return dest.doWrite(func(w io.Writer) error {
if err := car.WriteHeader(&car.CarHeader{
Roots: roots,
Version: 1,
}, w); err != nil {
return fmt.Errorf("failed to write car header: %s", err)
}
cs := cid.NewSet()
for _, dagSpec := range dags {
2022-10-29 14:11:09 +00:00
dagSpec := dagSpec
if err := utils.TraverseDag(
ctx,
ds,
root,
dagSpec.selector,
func(node format.Node) error {
if dagSpec.exportAll {
lk.Lock()
defer lk.Unlock()
if cs.Visit(node.Cid()) {
err := util.LdWrite(w, node.Cid().Bytes(), node.RawData())
if err != nil {
return xerrors.Errorf("writing block data: %w", err)
}
}
}
return nil
},
func(p traversal.Progress, n ipld.Node, r traversal.VisitReason) error {
if !dagSpec.exportAll && r == traversal.VisitReason_SelectionMatch {
var c cid.Cid
if p.LastBlock.Link == nil {
c = root
} else {
cidLnk, castOK := p.LastBlock.Link.(cidlink.Link)
if !castOK {
return xerrors.Errorf("cidlink cast unexpectedly failed on '%s'", p.LastBlock.Link)
}
c = cidLnk.Cid
}
if cs.Visit(c) {
2021-12-11 21:03:00 +00:00
nb, err := bs.Get(ctx, c)
if err != nil {
return xerrors.Errorf("getting block data: %w", err)
}
err = util.LdWrite(w, c.Bytes(), nb.RawData())
if err != nil {
return xerrors.Errorf("writing block data: %w", err)
}
}
return nil
}
return nil
},
); err != nil {
return xerrors.Errorf("error while traversing car dag: %w", err)
}
}
return nil
2021-11-11 16:17:11 +00:00
})
2021-11-09 16:27:42 +00:00
}
func (a *API) outputUnixFS(ctx context.Context, root cid.Cid, ds format.DAGService, dest ExportDest) error {
nd, err := ds.Get(ctx, root)
if err != nil {
return xerrors.Errorf("ClientRetrieve: %w", err)
}
file, err := unixfile.NewUnixfsFile(ctx, ds, nd)
if err != nil {
return xerrors.Errorf("ClientRetrieve: %w", err)
}
if dest.Writer == nil {
return files.WriteTo(file, dest.Path)
}
switch f := file.(type) {
case files.File:
_, err = io.Copy(dest.Writer, f)
if err != nil {
return err
}
return nil
default:
return fmt.Errorf("file type %T is not supported", nd)
}
}
2021-11-11 15:17:39 +00:00
type dagSpec struct {
root cid.Cid
selector ipld.Node
exportAll bool
2021-11-11 15:17:39 +00:00
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
func parseDagSpec(ctx context.Context, root cid.Cid, dsp []api.DagSpec, ds format.DAGService, car bool) ([]dagSpec, error) {
2021-11-11 15:17:39 +00:00
if len(dsp) == 0 {
return []dagSpec{
{
root: root,
selector: nil,
},
}, nil
}
2021-11-11 15:17:39 +00:00
out := make([]dagSpec, len(dsp))
for i, spec := range dsp {
out[i].exportAll = spec.ExportMerkleProof
if spec.DataSelector == nil {
return nil, xerrors.Errorf("invalid DagSpec at position %d: `DataSelector` can not be nil", i)
}
2021-11-23 16:42:43 +00:00
// reify selector
var err error
out[i].selector, err = getDataSelector(spec.DataSelector, car && spec.ExportMerkleProof)
if err != nil {
return nil, err
}
// find the pointed-at root node within the containing ds
var rsn ipld.Node
2021-11-11 15:17:39 +00:00
if strings.HasPrefix(string(*spec.DataSelector), "{") {
var err error
rsn, err = selectorparse.ParseJSONSelector(string(*spec.DataSelector))
if err != nil {
return nil, xerrors.Errorf("failed to parse json-selector '%s': %w", *spec.DataSelector, err)
}
} else {
selspec, _ := textselector.SelectorSpecFromPath(textselector.Expression(*spec.DataSelector), car && spec.ExportMerkleProof, nil) //nolint:errcheck
rsn = selspec.Node()
}
2021-11-11 15:17:39 +00:00
var newRoot cid.Cid
var errHalt = errors.New("halt walk")
if err := utils.TraverseDag(
ctx,
ds,
root,
rsn,
nil,
func(p traversal.Progress, n ipld.Node, r traversal.VisitReason) error {
if r == traversal.VisitReason_SelectionMatch {
if !car && p.LastBlock.Path.String() != p.Path.String() {
return xerrors.Errorf("unsupported selection path '%s' does not correspond to a block boundary (a.k.a. CID link)", p.Path.String())
}
2021-11-23 16:42:43 +00:00
if p.LastBlock.Link == nil {
// this is likely the root node that we've matched here
newRoot = root
2021-11-23 16:42:43 +00:00
return errHalt
}
cidLnk, castOK := p.LastBlock.Link.(cidlink.Link)
if !castOK {
return xerrors.Errorf("cidlink cast unexpectedly failed on '%s'", p.LastBlock.Link)
}
newRoot = cidLnk.Cid
2021-11-16 11:04:03 +00:00
return errHalt
}
return nil
},
); err != nil && err != errHalt {
return nil, xerrors.Errorf("error while locating partial retrieval sub-root: %w", err)
}
if newRoot == cid.Undef {
return nil, xerrors.Errorf("path selection does not match a node within %s", root)
}
out[i].root = newRoot
}
2021-11-11 15:17:39 +00:00
return out, nil
}
func (a *API) ClientListRetrievals(ctx context.Context) ([]api.RetrievalInfo, error) {
deals, err := a.Retrieval.ListDeals()
if err != nil {
return nil, err
}
dataTransfersByID, err := a.transfersByID(ctx)
if err != nil {
return nil, err
}
out := make([]api.RetrievalInfo, 0, len(deals))
for _, v := range deals {
// Find the data transfer associated with this deal
var transferCh *api.DataTransferChannel
if v.ChannelID != nil {
if ch, ok := dataTransfersByID[*v.ChannelID]; ok {
transferCh = &ch
}
}
out = append(out, a.newRetrievalInfoWithTransfer(transferCh, v))
}
sort.Slice(out, func(a, b int) bool {
return out[a].ID < out[b].ID
})
return out, nil
}
func (a *API) ClientGetRetrievalUpdates(ctx context.Context) (<-chan api.RetrievalInfo, error) {
updates := make(chan api.RetrievalInfo)
2021-11-11 15:17:39 +00:00
unsub := a.Retrieval.SubscribeToEvents(func(evt rm.ClientEvent, deal rm.ClientDealState) {
update := a.newRetrievalInfo(ctx, deal)
update.Event = &evt
select {
case updates <- update:
case <-ctx.Done():
}
})
go func() {
defer unsub()
<-ctx.Done()
}()
return updates, nil
}
func (a *API) newRetrievalInfoWithTransfer(ch *api.DataTransferChannel, deal rm.ClientDealState) api.RetrievalInfo {
return api.RetrievalInfo{
PayloadCID: deal.PayloadCID,
ID: deal.ID,
PieceCID: deal.PieceCID,
PricePerByte: deal.PricePerByte,
UnsealPrice: deal.UnsealPrice,
Status: deal.Status,
Message: deal.Message,
Provider: deal.Sender,
BytesReceived: deal.TotalReceived,
BytesPaidFor: deal.BytesPaidFor,
TotalPaid: deal.FundsSpent,
TransferChannelID: deal.ChannelID,
DataTransfer: ch,
}
}
func (a *API) newRetrievalInfo(ctx context.Context, v rm.ClientDealState) api.RetrievalInfo {
// Find the data transfer associated with this deal
var transferCh *api.DataTransferChannel
if v.ChannelID != nil {
state, err := a.DataTransfer.ChannelState(ctx, *v.ChannelID)
// Note: If there was an error just ignore it, as the data transfer may
// be not found if it's no longer active
if err == nil {
ch := api.NewDataTransferChannel(a.Host.ID(), state)
ch.Stages = state.Stages()
transferCh = &ch
}
}
return a.newRetrievalInfoWithTransfer(transferCh, v)
}
const dealProtoPrefix = "/fil/storage/mk/"
func (a *API) ClientQueryAsk(ctx context.Context, p peer.ID, miner address.Address) (*api.StorageAsk, error) {
2020-08-05 20:54:45 +00:00
mi, err := a.StateMinerInfo(ctx, miner, types.EmptyTSK)
if err != nil {
return nil, xerrors.Errorf("failed getting miner info: %w", err)
}
info := utils.NewStorageProviderInfo(miner, mi.Worker, mi.SectorSize, p, mi.Multiaddrs)
2020-09-29 11:53:30 +00:00
ask, err := a.SMDealClient.GetAsk(ctx, info)
if err != nil {
return nil, err
}
res := &api.StorageAsk{
Response: ask,
}
ps, err := a.Host.Peerstore().GetProtocols(p)
if err != nil {
return nil, err
}
for _, s := range ps {
if strings.HasPrefix(s, dealProtoPrefix) {
res.DealProtocols = append(res.DealProtocols, s)
}
}
sort.Strings(res.DealProtocols)
return res, nil
2019-09-13 21:00:36 +00:00
}
func (a *API) ClientCalcCommP(ctx context.Context, inpath string) (*api.CommPRet, error) {
rdr, err := os.Open(inpath)
if err != nil {
return nil, err
}
defer rdr.Close() //nolint:errcheck
2021-02-16 19:48:31 +00:00
// check that the data is a car file; if it's not, retrieval won't work
2021-12-11 21:03:00 +00:00
_, err = car.ReadHeader(bufio.NewReader(rdr))
2021-02-16 19:48:31 +00:00
if err != nil {
return nil, xerrors.Errorf("not a car file: %w", err)
}
if _, err := rdr.Seek(0, io.SeekStart); err != nil {
return nil, xerrors.Errorf("seek to start: %w", err)
}
w := &writer.Writer{}
_, err = io.CopyBuffer(w, rdr, make([]byte, writer.CommPBuf))
if err != nil {
return nil, xerrors.Errorf("copy into commp writer: %w", err)
}
commp, err := w.Sum()
if err != nil {
return nil, xerrors.Errorf("computing commP failed: %w", err)
}
return &api.CommPRet{
Root: commp.PieceCID,
Size: commp.PieceSize.Unpadded(),
}, nil
}
2020-07-31 16:22:04 +00:00
type lenWriter int64
func (w *lenWriter) Write(p []byte) (n int, err error) {
*w += lenWriter(len(p))
return len(p), nil
}
func (a *API) ClientDealSize(ctx context.Context, root cid.Cid) (api.DataSize, error) {
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
bs, onDone, err := a.dealBlockstore(root)
if err != nil {
return api.DataSize{}, err
}
defer onDone()
2020-07-31 16:22:04 +00:00
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
dag := merkledag.NewDAGService(blockservice.New(bs, offline.Exchange(bs)))
2020-07-31 16:22:04 +00:00
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
var w lenWriter
err = car.WriteCar(ctx, dag, []cid.Cid{root}, &w)
2020-07-31 16:22:04 +00:00
if err != nil {
return api.DataSize{}, err
}
up := padreader.PaddedSize(uint64(w))
return api.DataSize{
PayloadSize: int64(w),
PieceSize: up.Padded(),
}, nil
}
func (a *API) ClientDealPieceCID(ctx context.Context, root cid.Cid) (api.DataCIDSize, error) {
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
bs, onDone, err := a.dealBlockstore(root)
if err != nil {
return api.DataCIDSize{}, err
}
defer onDone()
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
dag := merkledag.NewDAGService(blockservice.New(bs, offline.Exchange(bs)))
w := &writer.Writer{}
bw := bufio.NewWriterSize(w, int(writer.CommPBuf))
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
err = car.WriteCar(ctx, dag, []cid.Cid{root}, w)
if err != nil {
return api.DataCIDSize{}, err
}
if err := bw.Flush(); err != nil {
return api.DataCIDSize{}, err
}
dataCIDSize, err := w.Sum()
return api.DataCIDSize(dataCIDSize), err
}
func (a *API) ClientGenCar(ctx context.Context, ref api.FileRef, outputPath string) error {
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
// create a temporary import to represent this job and obtain a staging CAR.
id, err := a.importManager().CreateImport()
2020-07-06 23:39:30 +00:00
if err != nil {
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
return xerrors.Errorf("failed to create temporary import: %w", err)
2020-07-06 23:39:30 +00:00
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
defer a.importManager().Remove(id) //nolint:errcheck
tmp, err := a.importManager().AllocateCAR(id)
if err != nil {
return xerrors.Errorf("failed to allocate temporary CAR: %w", err)
2020-07-06 23:39:30 +00:00
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
defer os.Remove(tmp) //nolint:errcheck
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
// generate and import the UnixFS DAG into a filestore (positional reference) CAR.
2022-03-21 09:37:35 +00:00
root, err := unixfs.CreateFilestore(ctx, ref.Path, tmp)
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
if err != nil {
return xerrors.Errorf("failed to import file using unixfs: %w", err)
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
// open the positional reference CAR as a filestore.
fs, err := stores.ReadOnlyFilestore(tmp)
if err != nil {
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
return xerrors.Errorf("failed to open filestore from carv2 in path %s: %w", tmp, err)
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
defer fs.Close() //nolint:errcheck
f, err := os.Create(outputPath)
if err != nil {
return err
}
2021-10-07 08:47:51 +00:00
// build a dense deterministic CAR (dense = containing filled leaves)
if err := car.NewSelectiveCar(
ctx,
fs,
[]car.Dag{{
Root: root,
Selector: selectorparse.CommonSelector_ExploreAllRecursively,
}},
car.MaxTraversalLinks(config.MaxTraversalLinks),
).Write(
f,
); err != nil {
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
return xerrors.Errorf("failed to write CAR to output file: %w", err)
}
return f.Close()
}
func (a *API) ClientListDataTransfers(ctx context.Context) ([]api.DataTransferChannel, error) {
inProgressChannels, err := a.DataTransfer.InProgressChannels(ctx)
if err != nil {
return nil, err
}
apiChannels := make([]api.DataTransferChannel, 0, len(inProgressChannels))
for _, channelState := range inProgressChannels {
apiChannels = append(apiChannels, api.NewDataTransferChannel(a.Host.ID(), channelState))
}
return apiChannels, nil
}
func (a *API) ClientDataTransferUpdates(ctx context.Context) (<-chan api.DataTransferChannel, error) {
channels := make(chan api.DataTransferChannel)
unsub := a.DataTransfer.SubscribeToEvents(func(evt datatransfer.Event, channelState datatransfer.ChannelState) {
channel := api.NewDataTransferChannel(a.Host.ID(), channelState)
select {
case <-ctx.Done():
case channels <- channel:
}
})
go func() {
defer unsub()
<-ctx.Done()
}()
return channels, nil
}
2020-08-27 18:32:51 +00:00
func (a *API) ClientRestartDataTransfer(ctx context.Context, transferID datatransfer.TransferID, otherPeer peer.ID, isInitiator bool) error {
selfPeer := a.Host.ID()
if isInitiator {
return a.DataTransfer.RestartDataTransferChannel(ctx, datatransfer.ChannelID{Initiator: selfPeer, Responder: otherPeer, ID: transferID})
}
return a.DataTransfer.RestartDataTransferChannel(ctx, datatransfer.ChannelID{Initiator: otherPeer, Responder: selfPeer, ID: transferID})
2020-10-22 20:40:26 +00:00
}
func (a *API) ClientCancelDataTransfer(ctx context.Context, transferID datatransfer.TransferID, otherPeer peer.ID, isInitiator bool) error {
selfPeer := a.Host.ID()
if isInitiator {
return a.DataTransfer.CloseDataTransferChannel(ctx, datatransfer.ChannelID{Initiator: selfPeer, Responder: otherPeer, ID: transferID})
}
return a.DataTransfer.CloseDataTransferChannel(ctx, datatransfer.ChannelID{Initiator: otherPeer, Responder: selfPeer, ID: transferID})
}
2020-09-04 05:34:59 +00:00
func (a *API) ClientRetrieveTryRestartInsufficientFunds(ctx context.Context, paymentChannel address.Address) error {
return a.Retrieval.TryRestartInsufficientFunds(paymentChannel)
}
func (a *API) ClientGetDealStatus(ctx context.Context, statusCode uint64) (string, error) {
ststr, ok := storagemarket.DealStates[statusCode]
if !ok {
return "", fmt.Errorf("no such deal state %d", statusCode)
}
return ststr, nil
}
integrate DAG store and CARv2 in deal-making (#6671) This commit removes badger from the deal-making processes, and moves to a new architecture with the dagstore as the cental component on the miner-side, and CARv2s on the client-side. Every deal that has been handed off to the sealing subsystem becomes a shard in the dagstore. Shards are mounted via the LotusMount, which teaches the dagstore how to load the related piece when serving retrievals. When the miner starts the Lotus for the first time with this patch, we will perform a one-time migration of all active deals into the dagstore. This is a lightweight process, and it consists simply of registering the shards in the dagstore. Shards are backed by the unsealed copy of the piece. This is currently a CARv1. However, the dagstore keeps CARv2 indices for all pieces, so when it's time to acquire a shard to serve a retrieval, the unsealed CARv1 is joined with its index (safeguarded by the dagstore), to form a read-only blockstore, thus taking the place of the monolithic badger. Data transfers have been adjusted to interface directly with CARv2 files. On inbound transfers (client retrievals, miner storage deals), we stream the received data into a CARv2 ReadWrite blockstore. On outbound transfers (client storage deals, miner retrievals), we serve the data off a CARv2 ReadOnly blockstore. Client-side imports are managed by the refactored *imports.Manager component (when not using IPFS integration). Just like it before, we use the go-filestore library to avoid duplicating the data from the original file in the resulting UnixFS DAG (concretely the leaves). However, the target of those imports are what we call "ref-CARv2s": CARv2 files placed under the `$LOTUS_PATH/imports` directory, containing the intermediate nodes in full, and the leaves as positional references to the original file on disk. Client-side retrievals are placed into CARv2 files in the location: `$LOTUS_PATH/retrievals`. A new set of `Dagstore*` JSON-RPC operations and `lotus-miner dagstore` subcommands have been introduced on the miner-side to inspect and manage the dagstore. Despite moving to a CARv2-backed system, the IPFS integration has been respected, and it continues to be possible to make storage deals with data held in an IPFS node, and to perform retrievals directly into an IPFS node. NOTE: because the "staging" and "client" Badger blockstores are no longer used, existing imports on the client will be rendered useless. On startup, Lotus will enumerate all imports and print WARN statements on the log for each import that needs to be reimported. These log lines contain these messages: - import lacks carv2 path; import will not work; please reimport - import has missing/broken carv2; please reimport At the end, we will print a "sanity check completed" message indicating the count of imports found, and how many were deemed broken. Co-authored-by: Aarsh Shah <aarshkshah1992@gmail.com> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com> Co-authored-by: Raúl Kripalani <raul@protocol.ai> Co-authored-by: Dirk McCormick <dirkmdev@gmail.com>
2021-08-16 22:34:32 +00:00
// dealBlockstore picks the source blockstore for a storage deal; either the
// IPFS blockstore, or an import CARv2 file. It also returns a function that
// must be called when done.
func (a *API) dealBlockstore(root cid.Cid) (bstore.Blockstore, func(), error) {
switch acc := a.StorageBlockstoreAccessor.(type) {
case *storageadapter.ImportsBlockstoreAccessor:
bs, err := acc.Get(root)
if err != nil {
return nil, nil, xerrors.Errorf("no import found for root %s: %w", root, err)
}
doneFn := func() {
_ = acc.Done(root) //nolint:errcheck
}
return bs, doneFn, nil
case *storageadapter.ProxyBlockstoreAccessor:
return acc.Blockstore, func() {}, nil
default:
return nil, nil, xerrors.Errorf("unsupported blockstore accessor type: %T", acc)
}
}