lotus/node/modules/client.go

232 lines
9.4 KiB
Go
Raw Normal View History

2019-08-01 14:19:53 +00:00
package modules
import (
"bytes"
2019-08-01 14:19:53 +00:00
"context"
"os"
"path/filepath"
"time"
2020-03-18 19:43:06 +00:00
"go.uber.org/fx"
2021-03-22 11:08:05 +00:00
"golang.org/x/xerrors"
2020-03-18 19:43:06 +00:00
2021-03-22 11:08:05 +00:00
"github.com/filecoin-project/go-data-transfer/channelmonitor"
dtimpl "github.com/filecoin-project/go-data-transfer/impl"
dtnet "github.com/filecoin-project/go-data-transfer/network"
dtgstransport "github.com/filecoin-project/go-data-transfer/transport/graphsync"
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-state-types/abi"
"github.com/ipfs/go-datastore"
"github.com/ipfs/go-datastore/namespace"
"github.com/libp2p/go-libp2p-core/host"
2020-09-29 11:53:30 +00:00
"github.com/filecoin-project/go-fil-markets/discovery"
discoveryimpl "github.com/filecoin-project/go-fil-markets/discovery/impl"
"github.com/filecoin-project/go-fil-markets/retrievalmarket"
retrievalimpl "github.com/filecoin-project/go-fil-markets/retrievalmarket/impl"
rmnet "github.com/filecoin-project/go-fil-markets/retrievalmarket/network"
"github.com/filecoin-project/go-fil-markets/storagemarket"
2020-01-14 01:24:04 +00:00
storageimpl "github.com/filecoin-project/go-fil-markets/storagemarket/impl"
2020-03-18 18:57:22 +00:00
"github.com/filecoin-project/go-fil-markets/storagemarket/impl/requestvalidation"
smnet "github.com/filecoin-project/go-fil-markets/storagemarket/network"
"github.com/filecoin-project/lotus/blockstore"
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/markets/storageadapter"
"github.com/filecoin-project/lotus/chain/market"
"github.com/filecoin-project/lotus/journal"
2020-09-14 18:01:35 +00:00
"github.com/filecoin-project/lotus/markets"
2020-07-31 01:27:42 +00:00
marketevents "github.com/filecoin-project/lotus/markets/loggers"
2020-01-14 01:24:04 +00:00
"github.com/filecoin-project/lotus/markets/retrievaladapter"
"github.com/filecoin-project/lotus/node/config"
2020-03-18 18:57:22 +00:00
"github.com/filecoin-project/lotus/node/impl/full"
payapi "github.com/filecoin-project/lotus/node/impl/paych"
"github.com/filecoin-project/lotus/node/modules/dtypes"
2021-12-17 09:42:09 +00:00
"github.com/filecoin-project/lotus/node/modules/helpers"
"github.com/filecoin-project/lotus/node/repo"
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/imports"
2019-08-01 14:19:53 +00:00
)
2021-12-17 09:42:09 +00:00
func HandleMigrateClientFunds(lc fx.Lifecycle, mctx helpers.MetricsCtx, ds dtypes.MetadataDS, wallet full.WalletAPI, fundMgr *market.FundManager) {
lc.Append(fx.Hook{
OnStart: func(ctx context.Context) error {
addr, err := wallet.WalletDefaultAddress(ctx)
// nothing to be done if there is no default address
if err != nil {
return nil
}
2021-12-17 09:42:09 +00:00
b, err := ds.Get(helpers.LifecycleCtx(mctx, lc), datastore.NewKey("/marketfunds/client"))
if err != nil {
if xerrors.Is(err, datastore.ErrNotFound) {
return nil
}
2021-02-11 11:00:26 +00:00
log.Errorf("client funds migration - getting datastore value: %v", err)
return nil
}
var value abi.TokenAmount
if err = value.UnmarshalCBOR(bytes.NewReader(b)); err != nil {
2021-02-11 11:00:26 +00:00
log.Errorf("client funds migration - unmarshalling datastore value: %v", err)
return nil
}
_, err = fundMgr.Reserve(ctx, addr, addr, value)
if err != nil {
2021-02-11 11:00:26 +00:00
log.Errorf("client funds migration - reserving funds (wallet %s, addr %s, funds %d): %v",
addr, addr, value, err)
return nil
}
2021-12-17 09:42:09 +00:00
return ds.Delete(helpers.LifecycleCtx(mctx, lc), datastore.NewKey("/marketfunds/client"))
},
})
}
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 ClientImportMgr(ds dtypes.MetadataDS, r repo.LockedRepo) (dtypes.ClientImportMgr, error) {
// store the imports under the repo's `imports` subdirectory.
dir := filepath.Join(r.Path(), "imports")
if err := os.MkdirAll(dir, 0755); err != nil {
return nil, xerrors.Errorf("failed to create directory %s: %w", dir, 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
ns := namespace.Wrap(ds, datastore.NewKey("/client"))
return imports.NewManager(ns, dir), nil
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
// TODO this should be removed.
func ClientBlockstore() dtypes.ClientBlockstore {
// in most cases this is now unused in normal operations -- however, it's important to preserve for the IPFS use case
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 blockstore.WrapIDStore(blockstore.FromDatastore(datastore.NewMapDatastore()))
2019-08-26 13:45:36 +00:00
}
2019-08-06 22:04:21 +00:00
// RegisterClientValidator is an initialization hook that registers the client
// request validator with the data transfer module as the validator for
// StorageDataTransferVoucher types
func RegisterClientValidator(crv dtypes.ClientRequestValidator, dtm dtypes.ClientDataTransfer) {
if err := dtm.RegisterVoucherType(&requestvalidation.StorageDataTransferVoucher{}, (*requestvalidation.UnifiedRequestValidator)(crv)); err != nil {
2019-12-05 05:02:14 +00:00
panic(err)
}
}
// NewClientGraphsyncDataTransfer returns a data transfer manager that just
// uses the clients's Client DAG service for transfers
func NewClientGraphsyncDataTransfer(lc fx.Lifecycle, h host.Host, gs dtypes.Graphsync, ds dtypes.MetadataDS, r repo.LockedRepo) (dtypes.ClientDataTransfer, error) {
// go-data-transfer protocol retries:
// 1s, 5s, 25s, 2m5s, 5m x 11 ~= 1 hour
dtRetryParams := dtnet.RetryParameters(time.Second, 5*time.Minute, 15, 5)
net := dtnet.NewFromLibp2pHost(h, dtRetryParams)
dtDs := namespace.Wrap(ds, datastore.NewKey("/datatransfer/client/transfers"))
2021-10-05 08:40:01 +00:00
transport := dtgstransport.NewTransport(h.ID(), gs, net)
err := os.MkdirAll(filepath.Join(r.Path(), "data-transfer"), 0755) //nolint: gosec
if err != nil && !os.IsExist(err) {
return nil, err
}
2021-03-22 11:08:05 +00:00
// data-transfer push / pull channel restart configuration:
dtRestartConfig := dtimpl.ChannelRestartConfig(channelmonitor.Config{
2021-06-01 16:16:12 +00:00
// Disable Accept and Complete timeouts until this issue is resolved:
// https://github.com/filecoin-project/lotus/issues/6343#
// Wait for the other side to respond to an Open channel message
AcceptTimeout: 0,
// Wait for the other side to send a Complete message once all
// data has been sent / received
CompleteTimeout: 0,
2021-04-23 13:32:50 +00:00
// When an error occurs, wait a little while until all related errors
// have fired before sending a restart message
RestartDebounce: 10 * time.Second,
2021-03-22 11:08:05 +00:00
// After sending a restart, wait for at least 1 minute before sending another
RestartBackoff: time.Minute,
// After trying to restart 3 times, give up and fail the transfer
MaxConsecutiveRestarts: 3,
})
dt, err := dtimpl.NewDataTransfer(dtDs, filepath.Join(r.Path(), "data-transfer"), net, transport, dtRestartConfig)
if err != nil {
return nil, err
}
dt.OnReady(marketevents.ReadyLogger("client data transfer"))
lc.Append(fx.Hook{
OnStart: func(ctx context.Context) error {
2020-11-23 09:31:39 +00:00
dt.SubscribeToEvents(marketevents.DataTransferLogger)
return dt.Start(ctx)
},
OnStop: func(ctx context.Context) error {
return dt.Stop(ctx)
},
})
return dt, nil
}
2020-03-18 18:57:22 +00:00
// NewClientDatastore creates a datastore for the client to store its deals
func NewClientDatastore(ds dtypes.MetadataDS) dtypes.ClientDatastore {
return namespace.Wrap(ds, datastore.NewKey("/deals/client"))
}
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
// StorageBlockstoreAccessor returns the default storage blockstore accessor
// from the import manager.
func StorageBlockstoreAccessor(importmgr dtypes.ClientImportMgr) storagemarket.BlockstoreAccessor {
return storageadapter.NewImportsBlockstoreAccessor(importmgr)
}
// RetrievalBlockstoreAccessor returns the default retrieval blockstore accessor
// using the subdirectory `retrievals` under the repo.
func RetrievalBlockstoreAccessor(r repo.LockedRepo) (retrievalmarket.BlockstoreAccessor, error) {
dir := filepath.Join(r.Path(), "retrievals")
if err := os.MkdirAll(dir, 0755); err != nil {
return nil, xerrors.Errorf("failed to create directory %s: %w", dir, err)
}
return retrievaladapter.NewCARBlockstoreAccessor(dir), nil
}
func StorageClient(lc fx.Lifecycle, h host.Host, dataTransfer dtypes.ClientDataTransfer, discovery *discoveryimpl.Local,
deals dtypes.ClientDatastore, scn storagemarket.StorageClientNode, accessor storagemarket.BlockstoreAccessor, j journal.Journal) (storagemarket.StorageClient, error) {
// go-fil-markets protocol retries:
// 1s, 5s, 25s, 2m5s, 5m x 11 ~= 1 hour
marketsRetryParams := smnet.RetryParameters(time.Second, 5*time.Minute, 15, 5)
net := smnet.NewFromLibp2pHost(h, marketsRetryParams)
c, err := storageimpl.NewClient(net, dataTransfer, discovery, deals, scn, accessor, storageimpl.DealPollingInterval(time.Second), storageimpl.MaxTraversalLinks(config.MaxTraversalLinks))
if err != nil {
return nil, err
}
2020-09-29 11:53:30 +00:00
c.OnReady(marketevents.ReadyLogger("storage client"))
lc.Append(fx.Hook{
OnStart: func(ctx context.Context) error {
2020-07-31 01:27:42 +00:00
c.SubscribeToEvents(marketevents.StorageClientLogger)
evtType := j.RegisterEventType("markets/storage/client", "state_change")
c.SubscribeToEvents(markets.StorageClientJournaler(j, evtType))
return c.Start(ctx)
},
OnStop: func(context.Context) error {
return c.Stop()
},
})
return c, nil
}
// RetrievalClient creates a new retrieval client attached to the client blockstore
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 RetrievalClient(lc fx.Lifecycle, h host.Host, r repo.LockedRepo, dt dtypes.ClientDataTransfer, payAPI payapi.PaychAPI, resolver discovery.PeerResolver,
ds dtypes.MetadataDS, chainAPI full.ChainAPI, stateAPI full.StateAPI, accessor retrievalmarket.BlockstoreAccessor, j journal.Journal) (retrievalmarket.RetrievalClient, error) {
2020-09-04 05:34:59 +00:00
adapter := retrievaladapter.NewRetrievalClientNode(payAPI, chainAPI, stateAPI)
network := rmnet.NewFromLibp2pHost(h)
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
ds = namespace.Wrap(ds, datastore.NewKey("/retrievals/client"))
client, err := retrievalimpl.NewClient(network, dt, adapter, resolver, ds, accessor)
2020-07-31 01:27:42 +00:00
if err != nil {
return nil, err
}
2020-09-29 11:53:30 +00:00
client.OnReady(marketevents.ReadyLogger("retrieval client"))
2020-07-31 01:27:42 +00:00
lc.Append(fx.Hook{
OnStart: func(ctx context.Context) error {
client.SubscribeToEvents(marketevents.RetrievalClientLogger)
evtType := j.RegisterEventType("markets/retrieval/client", "state_change")
client.SubscribeToEvents(markets.RetrievalClientJournaler(j, evtType))
2020-09-29 11:53:30 +00:00
return client.Start(ctx)
2020-07-31 01:27:42 +00:00
},
})
return client, nil
}