ipld-eth-server/pkg/seed_node/service.go

424 lines
14 KiB
Go

// VulcanizeDB
// Copyright © 2019 Vulcanize
// This program is free software: you can redistribute it and/or modify
// it under the terms of the GNU Affero General Public License as published by
// the Free Software Foundation, either version 3 of the License, or
// (at your option) any later version.
// This program is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU Affero General Public License for more details.
// You should have received a copy of the GNU Affero General Public License
// along with this program. If not, see <http://www.gnu.org/licenses/>.
package seed_node
import (
"sync"
"github.com/ethereum/go-ethereum/common"
"github.com/ethereum/go-ethereum/crypto"
"github.com/ethereum/go-ethereum/node"
"github.com/ethereum/go-ethereum/p2p"
"github.com/ethereum/go-ethereum/params"
"github.com/ethereum/go-ethereum/rlp"
"github.com/ethereum/go-ethereum/rpc"
"github.com/ethereum/go-ethereum/statediff"
log "github.com/sirupsen/logrus"
"github.com/vulcanize/vulcanizedb/libraries/shared/streamer"
"github.com/vulcanize/vulcanizedb/pkg/config"
"github.com/vulcanize/vulcanizedb/pkg/core"
"github.com/vulcanize/vulcanizedb/pkg/datastore/postgres"
"github.com/vulcanize/vulcanizedb/pkg/ipfs"
)
const payloadChanBufferSize = 20000 // the max eth sub buffer size
// NodeInterface is the top level interface for streaming, converting to IPLDs, publishing,
// and indexing all Ethereum data; screening this data; and serving it up to subscribed clients
// This service is compatible with the Ethereum service interface (node.Service)
type NodeInterface interface {
// APIs(), Protocols(), Start() and Stop()
node.Service
// Main event loop for syncAndPublish processes
SyncAndPublish(wg *sync.WaitGroup, forwardPayloadChan chan<- ipfs.IPLDPayload, forwardQuitchan chan<- bool) error
// Main event loop for handling client pub-sub
ScreenAndServe(screenAndServePayload <-chan ipfs.IPLDPayload, screenAndServeQuit <-chan bool)
// Method to subscribe to receive state diff processing output
Subscribe(id rpc.ID, sub chan<- streamer.SeedNodePayload, quitChan chan<- bool, streamFilters config.Subscription)
// Method to unsubscribe from state diff processing
Unsubscribe(id rpc.ID)
// Method to access the Geth node info for this service
Node() core.Node
}
// Service is the underlying struct for the SyncAndPublish interface
type Service struct {
// Used to sync access to the Subscriptions
sync.Mutex
// Interface for streaming statediff payloads over a geth rpc subscription
Streamer streamer.IStateDiffStreamer
// Interface for converting statediff payloads into ETH-IPLD object payloads
Converter ipfs.PayloadConverter
// Interface for publishing the ETH-IPLD payloads to IPFS
Publisher ipfs.IPLDPublisher
// Interface for indexing the CIDs of the published ETH-IPLDs in Postgres
Repository CIDRepository
// Interface for filtering and serving data according to subscribed clients according to their specification
Filterer ResponseFilterer
// Interface for fetching ETH-IPLD objects from IPFS
Fetcher ipfs.IPLDFetcher
// Interface for searching and retrieving CIDs from Postgres index
Retriever CIDRetriever
// Interface for resolving ipfs blocks to their data types
Resolver ipfs.IPLDResolver
// Chan the processor uses to subscribe to state diff payloads from the Streamer
PayloadChan chan statediff.Payload
// Used to signal shutdown of the service
QuitChan chan bool
// A mapping of rpc.IDs to their subscription channels, mapped to their subscription type (hash of the StreamFilters)
Subscriptions map[common.Hash]map[rpc.ID]Subscription
// A mapping of subscription hash type to the corresponding StreamFilters
SubscriptionTypes map[common.Hash]config.Subscription
// Number of workers
WorkerPoolSize int
// Info for the Geth node that this seed node is working with
gethNode core.Node
}
// NewSeedNode creates a new seed_node.Interface using an underlying seed_node.Service struct
func NewSeedNode(ipfsPath string, db *postgres.DB, rpcClient core.RpcClient, qc chan bool, workers int, node core.Node) (NodeInterface, error) {
publisher, err := ipfs.NewIPLDPublisher(ipfsPath)
if err != nil {
return nil, err
}
fetcher, err := ipfs.NewIPLDFetcher(ipfsPath)
if err != nil {
return nil, err
}
return &Service{
Streamer: streamer.NewStateDiffStreamer(rpcClient),
Repository: NewCIDRepository(db),
Converter: ipfs.NewPayloadConverter(params.MainnetChainConfig),
Publisher: publisher,
Filterer: NewResponseFilterer(),
Fetcher: fetcher,
Retriever: NewCIDRetriever(db),
Resolver: ipfs.NewIPLDResolver(),
PayloadChan: make(chan statediff.Payload, payloadChanBufferSize),
QuitChan: qc,
Subscriptions: make(map[common.Hash]map[rpc.ID]Subscription),
SubscriptionTypes: make(map[common.Hash]config.Subscription),
WorkerPoolSize: workers,
gethNode: node,
}, nil
}
// Protocols exports the services p2p protocols, this service has none
func (sap *Service) Protocols() []p2p.Protocol {
return []p2p.Protocol{}
}
// APIs returns the RPC descriptors the StateDiffingService offers
func (sap *Service) APIs() []rpc.API {
return []rpc.API{
{
Namespace: APIName,
Version: APIVersion,
Service: NewPublicSeedNodeAPI(sap),
Public: true,
},
}
}
// SyncAndPublish is the backend processing loop which streams data from geth, converts it to iplds, publishes them to ipfs, and indexes their cids
// This continues on no matter if or how many subscribers there are, it then forwards the data to the ScreenAndServe() loop
// which filters and sends relevant data to client subscriptions, if there are any
func (sap *Service) SyncAndPublish(wg *sync.WaitGroup, screenAndServePayload chan<- ipfs.IPLDPayload, screenAndServeQuit chan<- bool) error {
sub, err := sap.Streamer.Stream(sap.PayloadChan)
if err != nil {
return err
}
wg.Add(1)
// Channels for forwarding data to the publishAndIndex workers
publishAndIndexPayload := make(chan ipfs.IPLDPayload, payloadChanBufferSize)
publishAndIndexQuit := make(chan bool, sap.WorkerPoolSize)
// publishAndIndex worker pool to handle publishing and indexing concurrently, while
// limiting the number of Postgres connections we can possibly open so as to prevent error
for i := 0; i < sap.WorkerPoolSize; i++ {
sap.publishAndIndex(i, publishAndIndexPayload, publishAndIndexQuit)
}
go func() {
for {
select {
case payload := <-sap.PayloadChan:
ipldPayload, err := sap.Converter.Convert(payload)
if err != nil {
log.Error(err)
continue
}
// If we have a ScreenAndServe process running, forward the payload to it
select {
case screenAndServePayload <- *ipldPayload:
default:
}
// Forward the payload to the publishAndIndex workers
select {
case publishAndIndexPayload <- *ipldPayload:
default:
}
case err = <-sub.Err():
log.Error(err)
case <-sap.QuitChan:
// If we have a ScreenAndServe process running, forward the quit signal to it
select {
case screenAndServeQuit <- true:
default:
}
// Also forward a quit signal for each of the workers
for i := 0; i < sap.WorkerPoolSize; i++ {
select {
case publishAndIndexQuit <- true:
default:
}
}
log.Info("quiting SyncAndPublish process")
wg.Done()
return
}
}
}()
return nil
}
func (sap *Service) publishAndIndex(id int, publishAndIndexPayload <-chan ipfs.IPLDPayload, publishAndIndexQuit <-chan bool) {
go func() {
for {
select {
case payload := <-publishAndIndexPayload:
cidPayload, err := sap.Publisher.Publish(&payload)
if err != nil {
log.Errorf("worker %d error: %v", id, err)
continue
}
err = sap.Repository.Index(cidPayload)
if err != nil {
log.Errorf("worker %d error: %v", id, err)
}
case <-publishAndIndexQuit:
log.Infof("quiting publishAndIndex worker %d", id)
return
}
}
}()
}
// ScreenAndServe is the loop used to screen data streamed from the state diffing eth node
// and send the appropriate portions of it to a requesting client subscription, according to their subscription configuration
func (sap *Service) ScreenAndServe(screenAndServePayload <-chan ipfs.IPLDPayload, screenAndServeQuit <-chan bool) {
go func() {
for {
select {
case payload := <-screenAndServePayload:
err := sap.sendResponse(payload)
if err != nil {
log.Error(err)
}
case <-screenAndServeQuit:
log.Info("quiting ScreenAndServe process")
return
}
}
}()
}
func (sap *Service) sendResponse(payload ipfs.IPLDPayload) error {
sap.Lock()
for ty, subs := range sap.Subscriptions {
// Retrieve the subscription parameters for this subscription type
subConfig, ok := sap.SubscriptionTypes[ty]
if !ok {
log.Errorf("subscription configuration for subscription type %s not available", ty.Hex())
continue
}
response, err := sap.Filterer.FilterResponse(subConfig, payload)
if err != nil {
log.Error(err)
continue
}
for id, sub := range subs {
select {
case sub.PayloadChan <- response:
log.Infof("sending seed node payload to subscription %s", id)
default:
log.Infof("unable to send payload to subscription %s; channel has no receiver", id)
}
}
}
sap.Unlock()
return nil
}
// Subscribe is used by the API to subscribe to the service loop
func (sap *Service) Subscribe(id rpc.ID, sub chan<- streamer.SeedNodePayload, quitChan chan<- bool, streamFilters config.Subscription) {
log.Info("Subscribing to the seed node service")
// Subscription type is defined as the hash of its content
// Group subscriptions by type and screen payloads once for subs of the same type
by, err := rlp.EncodeToBytes(streamFilters)
if err != nil {
log.Error(err)
}
subscriptionHash := crypto.Keccak256(by)
subscriptionType := common.BytesToHash(subscriptionHash)
subscription := Subscription{
PayloadChan: sub,
QuitChan: quitChan,
}
// If the subscription requests a backfill, use the Postgres index to lookup and retrieve historical data
// Otherwise we only filter new data as it is streamed in from the state diffing geth node
if streamFilters.BackFill || streamFilters.BackFillOnly {
sap.backFill(subscription, id, streamFilters)
}
if !streamFilters.BackFillOnly {
sap.Lock()
if sap.Subscriptions[subscriptionType] == nil {
sap.Subscriptions[subscriptionType] = make(map[rpc.ID]Subscription)
}
sap.Subscriptions[subscriptionType][id] = subscription
sap.SubscriptionTypes[subscriptionType] = streamFilters
sap.Unlock()
}
}
func (sap *Service) backFill(sub Subscription, id rpc.ID, con config.Subscription) {
log.Debug("back-filling data for id", id)
// Retrieve cached CIDs relevant to this subscriber
var endingBlock int64
var startingBlock int64
var err error
startingBlock, err = sap.Retriever.RetrieveFirstBlockNumber()
if err != nil {
sub.PayloadChan <- streamer.SeedNodePayload{
ErrMsg: "unable to set block range start; error: " + err.Error(),
}
}
if startingBlock < con.StartingBlock.Int64() {
startingBlock = con.StartingBlock.Int64()
}
endingBlock, err = sap.Retriever.RetrieveLastBlockNumber()
if err != nil {
sub.PayloadChan <- streamer.SeedNodePayload{
ErrMsg: "unable to set block range end; error: " + err.Error(),
}
}
if endingBlock > con.EndingBlock.Int64() && con.EndingBlock.Int64() > 0 && con.EndingBlock.Int64() > startingBlock {
endingBlock = con.EndingBlock.Int64()
}
log.Debug("backfill starting block:", con.StartingBlock)
log.Debug("backfill ending block:", endingBlock)
// Backfilled payloads are sent concurrently to the streamed payloads, so the receiver needs to pay attention to
// the blocknumbers in the payloads they receive to keep things in order
// TODO: separate backfill into a different rpc subscription method altogether?
go func() {
for i := con.StartingBlock.Int64(); i <= endingBlock; i++ {
cidWrapper, err := sap.Retriever.RetrieveCIDs(con, i)
if err != nil {
sub.PayloadChan <- streamer.SeedNodePayload{
ErrMsg: "CID retrieval error: " + err.Error(),
}
continue
}
if ipfs.EmptyCIDWrapper(*cidWrapper) {
continue
}
blocksWrapper, err := sap.Fetcher.FetchCIDs(*cidWrapper)
if err != nil {
log.Error(err)
sub.PayloadChan <- streamer.SeedNodePayload{
ErrMsg: "IPLD fetching error: " + err.Error(),
}
continue
}
backFillIplds, err := sap.Resolver.ResolveIPLDs(*blocksWrapper)
if err != nil {
log.Error(err)
sub.PayloadChan <- streamer.SeedNodePayload{
ErrMsg: "IPLD resolving error: " + err.Error(),
}
continue
}
select {
case sub.PayloadChan <- backFillIplds:
log.Infof("sending seed node back-fill payload to subscription %s", id)
default:
log.Infof("unable to send back-fill payload to subscription %s; channel has no receiver", id)
}
}
}()
}
// Unsubscribe is used to unsubscribe to the StateDiffingService loop
func (sap *Service) Unsubscribe(id rpc.ID) {
log.Info("Unsubscribing from the seed node service")
sap.Lock()
for ty := range sap.Subscriptions {
delete(sap.Subscriptions[ty], id)
if len(sap.Subscriptions[ty]) == 0 {
// If we removed the last subscription of this type, remove the subscription type outright
delete(sap.Subscriptions, ty)
delete(sap.SubscriptionTypes, ty)
}
}
sap.Unlock()
}
// Start is used to begin the service
func (sap *Service) Start(*p2p.Server) error {
log.Info("Starting seed node service")
wg := new(sync.WaitGroup)
payloadChan := make(chan ipfs.IPLDPayload, payloadChanBufferSize)
quitChan := make(chan bool, 1)
if err := sap.SyncAndPublish(wg, payloadChan, quitChan); err != nil {
return err
}
sap.ScreenAndServe(payloadChan, quitChan)
return nil
}
// Stop is used to close down the service
func (sap *Service) Stop() error {
log.Info("Stopping seed node service")
close(sap.QuitChan)
return nil
}
// Node returns the Geth node info for this service
func (sap *Service) Node() core.Node {
return sap.gethNode
}
// close is used to close all listening subscriptions
func (sap *Service) close() {
sap.Lock()
for ty, subs := range sap.Subscriptions {
for id, sub := range subs {
select {
case sub.QuitChan <- true:
log.Infof("closing subscription %s", id)
default:
log.Infof("unable to close subscription %s; channel has no receiver", id)
}
}
delete(sap.Subscriptions, ty)
delete(sap.SubscriptionTypes, ty)
}
sap.Unlock()
}