Implement first iteration of node

The coordinator implementation has been refactored to allow all the goroutines
to be handled from the node.
This commit is contained in:
Eduard S
2020-09-08 11:31:51 +02:00
parent bcd93fee36
commit 475614cc3a
15 changed files with 759 additions and 160 deletions

View File

@@ -1,7 +1,7 @@
package coordinator
import (
"time"
"fmt"
ethCommon "github.com/ethereum/go-ethereum/common"
"github.com/hermeznetwork/hermez-node/batchbuilder"
@@ -14,25 +14,23 @@ import (
"github.com/iden3/go-merkletree/db/memory"
)
// ErrStop is returned when the function is stopped asynchronously via the stop
// channel. It doesn't indicate an error.
var ErrStop = fmt.Errorf("Stopped")
// Config contains the Coordinator configuration
type Config struct {
ForgerAddress ethCommon.Address
LoopInterval time.Duration
}
// Coordinator implements the Coordinator type
type Coordinator struct {
// m sync.Mutex
stopch chan bool
stopforgerch chan bool
forging bool
isForgeSeq bool // WIP just for testing while implementing
config Config
batchNum common.BatchNum
batchQueue *BatchQueue
serverProofPool ServerProofPool
// synchronizer *synchronizer.Synchronizer
@@ -61,123 +59,102 @@ func NewCoordinator(conf Config,
return &c
}
// Stop stops the Coordinator
func (c *Coordinator) Stop() {
log.Info("Stopping Coordinator")
c.stopch <- true
}
// Start starts the Coordinator service
func (c *Coordinator) Start() {
c.stopch = make(chan bool) // initialize channel
go func() {
log.Info("Starting Coordinator")
for {
select {
case <-c.stopch:
close(c.stopforgerch)
log.Info("Coordinator stopped")
return
case <-time.After(c.config.LoopInterval):
if !c.isForgeSequence() {
if c.forging {
log.Info("forging stopped")
c.forging = false
close(c.stopforgerch)
}
log.Debug("not in forge time")
continue
}
if !c.forging {
log.Info("Start Forging")
// c.batchNum = c.hdb.GetLastBatchNum() // uncomment when HistoryDB is ready
err := c.txsel.Reset(c.batchNum)
if err != nil {
log.Error("forging err: ", err)
}
err = c.batchBuilder.Reset(c.batchNum, true)
if err != nil {
log.Error("forging err: ", err)
}
c.batchQueue = NewBatchQueue()
c.forgerLoop()
c.forging = true
}
}
// ForgeLoopFn is the function ran in a loop that checks if it's time to forge
// and forges a batch if so and sends it to outBatchCh. Returns true if it's
// the coordinator turn to forge.
func (c *Coordinator) ForgeLoopFn(outBatchCh chan *BatchInfo, stopCh chan bool) (bool, error) {
if !c.isForgeSequence() {
if c.forging {
log.Info("stop forging")
c.forging = false
}
}()
}
// forgerLoop trigers goroutines for:
// - forgeSequence
// - proveSequence
// - forgeConfirmationSequence
func (c *Coordinator) forgerLoop() {
c.stopforgerch = make(chan bool) // initialize channel
go func() {
log.Info("forgeSequence started")
for {
select {
case <-c.stopforgerch:
log.Info("forgeSequence stopped")
return
case <-time.After(c.config.LoopInterval):
if err := c.forgeSequence(); err != nil {
log.Error("forgeSequence err: ", err)
}
}
log.Debug("not in forge time")
return false, nil
}
log.Debug("forge time")
if !c.forging {
log.Info("start forging")
// c.batchNum = c.hdb.GetLastBatchNum() // uncomment when HistoryDB is ready
err := c.txsel.Reset(c.batchNum)
if err != nil {
log.Errorw("TxSelector.Reset", "error", err)
return true, err
}
}()
go func() {
log.Info("proveSequence started")
for {
select {
case <-c.stopforgerch:
log.Info("proveSequence stopped")
return
case <-time.After(c.config.LoopInterval):
if err := c.proveSequence(); err != nil && err != common.ErrBatchQueueEmpty {
log.Error("proveSequence err: ", err)
}
}
err = c.batchBuilder.Reset(c.batchNum, true)
if err != nil {
log.Errorw("BatchBuilder.Reset", "error", err)
return true, err
}
}()
go func() {
log.Info("forgeConfirmationSequence started")
for {
select {
case <-c.stopforgerch:
log.Info("forgeConfirmationSequence stopped")
return
case <-time.After(c.config.LoopInterval):
if err := c.forgeConfirmationSequence(); err != nil {
log.Error("forgeConfirmationSequence err: ", err)
}
}
}
}()
}
// forgeSequence
func (c *Coordinator) forgeSequence() error {
// c.batchQueue = NewBatchQueue()
c.forging = true
}
// TODO once synchronizer has this method ready:
// If there's been a reorg, handle it
// handleReorg() function decides if the reorg must restart the pipeline or not
// if c.synchronizer.Reorg():
_ = c.handleReorg()
// 0. If there's an available server proof: Start pipeline for batchNum = batchNum + 1
serverProofInfo, err := c.serverProofPool.GetNextAvailable() // blocking call, returns when a server proof is available
// 0. If there's an available server proof: Start pipeline for batchNum = batchNum + 1.
// non-blocking call, returns nil if a server proof is
// not available, or non-nil otherwise.
serverProofInfo, err := c.serverProofPool.GetNextAvailable(stopCh)
if err != nil {
return err
return true, err
}
// remove transactions from the pool that have been there for too long
err = c.purgeRemoveByTimeout()
log.Debugw("start forge")
batchInfo, err := c.forge(serverProofInfo)
if err != nil {
return err
log.Errorw("forge", "error", err)
return true, err
}
log.Debugw("end forge", "batchNum", batchInfo.batchNum)
outBatchCh <- batchInfo
return true, nil
}
// GetProofCallForgeLoopFn is the function ran in a loop that gets a forged
// batch via inBatchCh, waits for the proof server to finish, calls the ForgeBatch
// function in the Rollup Smart Contract, and sends the batch to outBatchCh.
func (c *Coordinator) GetProofCallForgeLoopFn(inBatchCh, outBatchCh chan *BatchInfo, stopCh chan bool) error {
select {
case <-stopCh:
log.Info("forgeLoopFn stopped")
return ErrStop
case batchInfo := <-inBatchCh:
log.Debugw("start getProofCallForge", "batchNum", batchInfo.batchNum)
if err := c.getProofCallForge(batchInfo); err != nil {
return err
}
log.Debugw("end getProofCallForge", "batchNum", batchInfo.batchNum)
outBatchCh <- batchInfo
}
return nil
}
// ForgeCallConfirmLoopFn is the function ran in a loop that gets a batch that
// has been sent to the Rollup Smart Contract via inBatchCh and waits for the
// ethereum transaction confirmation.
func (c *Coordinator) ForgeCallConfirmLoopFn(inBatchCh chan *BatchInfo, stopCh chan bool) error {
select {
case <-stopCh:
log.Info("forgeConfirmLoopFn stopped")
return ErrStop
case batchInfo := <-inBatchCh:
log.Debugw("start forgeCallConfirm", "batchNum", batchInfo.batchNum)
if err := c.forgeCallConfirm(batchInfo); err != nil {
return err
}
log.Debugw("end forgeCallConfirm", "batchNum", batchInfo.batchNum)
}
return nil
}
func (c *Coordinator) forge(serverProofInfo *ServerProofInfo) (*BatchInfo, error) {
// remove transactions from the pool that have been there for too long
err := c.purgeRemoveByTimeout()
if err != nil {
return nil, err
}
c.batchNum = c.batchNum + 1
@@ -193,13 +170,13 @@ func (c *Coordinator) forgeSequence() error {
var l1UserTxs []*common.L1Tx = nil // tmp, depends on HistoryDB
l1UserTxsExtra, l1OperatorTxs, poolL2Txs, err = c.txsel.GetL1L2TxSelection(c.batchNum, l1UserTxs) // TODO once feesInfo is added to method return, add the var
if err != nil {
return err
return nil, err
}
} else {
// 2b: only L2 txs
poolL2Txs, err = c.txsel.GetL2TxSelection(c.batchNum) // TODO once feesInfo is added to method return, add the var
if err != nil {
return err
return nil, err
}
l1UserTxsExtra = nil
l1OperatorTxs = nil
@@ -211,7 +188,7 @@ func (c *Coordinator) forgeSequence() error {
// all the nonces smaller than the current one)
err = c.purgeInvalidDueToL2TxsSelection(poolL2Txs)
if err != nil {
return err
return nil, err
}
// 3. Save metadata from TxSelector output for BatchNum
@@ -224,31 +201,23 @@ func (c *Coordinator) forgeSequence() error {
l2Txs := common.PoolL2TxsToL2Txs(poolL2Txs)
zkInputs, err := c.batchBuilder.BuildBatch(configBatch, l1UserTxsExtra, l1OperatorTxs, l2Txs, nil) // TODO []common.TokenID --> feesInfo
if err != nil {
return err
return nil, err
}
// 5. Save metadata from BatchBuilder output for BatchNum
batchInfo.SetZKInputs(zkInputs)
log.Debugf("Batch builded, batchNum: %d ", c.batchNum)
// 6. Call an idle server proof with BatchBuilder output, save server proof info for batchNum
err = batchInfo.serverProof.CalculateProof(zkInputs)
if err != nil {
return err
return nil, err
}
c.batchQueue.Push(&batchInfo)
return nil
return &batchInfo, nil
}
// proveSequence gets the generated zkProof & sends it to the SmartContract
func (c *Coordinator) proveSequence() error {
batchInfo := c.batchQueue.Pop()
if batchInfo == nil {
// no batches in queue, return
log.Debug("not batch to prove yet")
return common.ErrBatchQueueEmpty
}
// getProofCallForge gets the generated zkProof & sends it to the SmartContract
func (c *Coordinator) getProofCallForge(batchInfo *BatchInfo) error {
serverProofInfo := batchInfo.serverProof
proof, err := serverProofInfo.GetProof() // blocking call, until not resolved don't continue. Returns when the proof server has calculated the proof
if err != nil {
@@ -276,7 +245,7 @@ func (c *Coordinator) proveSequence() error {
return nil
}
func (c *Coordinator) forgeConfirmationSequence() error {
func (c *Coordinator) forgeCallConfirm(batchInfo *BatchInfo) error {
// TODO strategy of this sequence TBD
// confirm eth txs and mark them as accepted sequence
// ethTx := ethTxStore.GetFirstPending()