gaspersic/go-ethereum
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

cmd, eth, miner: make recommit configurable (#17444)

Browse Source 
* cmd, eth, miner: make recommit configurable

* cmd, eth, les, miner: polish a bit

* miner: filter duplicate sealing work

* cmd: remove uncessary conversion

* miner: avoid microptimization in favor of cleaner code
This commit is contained in:
gary rong
2018-08-22 03:56:54 +08:00
committed by Péter Szilágyi
parent 522cfc68ff
commit b2c644ffb5
12 changed files with 360 additions and 71 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
miner/miner.go
View File

@@ -20,6 +20,7 @@ package miner
import (
"fmt"
"sync/atomic"
"time"
"github.com/ethereum/go-ethereum/common"
"github.com/ethereum/go-ethereum/consensus"
@@ -51,13 +52,13 @@ type Miner struct {
shouldStart int32 // should start indicates whether we should start after sync
}
func New(eth Backend, config *params.ChainConfig, mux *event.TypeMux, engine consensus.Engine) *Miner {
func New(eth Backend, config *params.ChainConfig, mux *event.TypeMux, engine consensus.Engine, recommit time.Duration) *Miner {
miner := &Miner{
eth: eth,
mux: mux,
engine: engine,
exitCh: make(chan struct{}),
worker: newWorker(config, engine, eth, mux),
worker: newWorker(config, engine, eth, mux, recommit),
canStart: 1,
}
go miner.update()
@@ -144,6 +145,11 @@ func (self *Miner) SetExtra(extra []byte) error {
return nil
}
// SetRecommitInterval sets the interval for sealing work resubmitting.
func (self *Miner) SetRecommitInterval(interval time.Duration) {
self.worker.setRecommitInterval(interval)
}
// Pending returns the currently pending block and associated state.
func (self *Miner) Pending() (*types.Block, *state.StateDB) {
return self.worker.pending()

202
miner/worker.go
View File

@@ -51,12 +51,27 @@ const (
// chainSideChanSize is the size of channel listening to ChainSideEvent.
chainSideChanSize = 10
// resubmitAdjustChanSize is the size of resubmitting interval adjustment channel.
resubmitAdjustChanSize = 10
// miningLogAtDepth is the number of confirmations before logging successful mining.
miningLogAtDepth = 5
// blockRecommitInterval is the time interval to recreate the mining block with
// minRecommitInterval is the minimal time interval to recreate the mining block with
// any newly arrived transactions.
blockRecommitInterval = 3 * time.Second
minRecommitInterval = 1 * time.Second
// maxRecommitInterval is the maximum time interval to recreate the mining block with
// any newly arrived transactions.
maxRecommitInterval = 15 * time.Second
// intervalAdjustRatio is the impact a single interval adjustment has on sealing work
// resubmitting interval.
intervalAdjustRatio = 0.1
// intervalAdjustBias is applied during the new resubmit interval calculation in favor of
// increasing upper limit or decreasing lower limit so that the limit can be reachable.
intervalAdjustBias = 200 * 1000.0 * 1000.0
)
// environment is the worker's current environment and holds all of the current state information.
@@ -89,11 +104,18 @@ const (
commitInterruptResubmit
)
// newWorkReq represents a request for new sealing work submitting with relative interrupt notifier.
type newWorkReq struct {
interrupt *int32
noempty bool
}
// intervalAdjust represents a resubmitting interval adjustment.
type intervalAdjust struct {
ratio float64
inc bool
}
// worker is the main object which takes care of submitting new work to consensus engine
// and gathering the sealing result.
type worker struct {
@@ -112,11 +134,13 @@ type worker struct {
chainSideSub event.Subscription
// Channels
newWorkCh chan *newWorkReq
taskCh chan *task
resultCh chan *task
startCh chan struct{}
exitCh chan struct{}
newWorkCh chan *newWorkReq
taskCh chan *task
resultCh chan *task
startCh chan struct{}
exitCh chan struct{}
resubmitIntervalCh chan time.Duration
resubmitAdjustCh chan *intervalAdjust
current *environment // An environment for current running cycle.
possibleUncles map[common.Hash]*types.Block // A set of side blocks as the possible uncle blocks.
@@ -132,30 +156,34 @@ type worker struct {
// atomic status counters
running int32 // The indicator whether the consensus engine is running or not.
newTxs int32 // New arrival transaction count since last sealing work submitting.
// Test hooks
newTaskHook func(*task) // Method to call upon receiving a new sealing task
skipSealHook func(*task) bool // Method to decide whether skipping the sealing.
fullTaskHook func() // Method to call before pushing the full sealing task
newTaskHook func(*task) // Method to call upon receiving a new sealing task.
skipSealHook func(*task) bool // Method to decide whether skipping the sealing.
fullTaskHook func() // Method to call before pushing the full sealing task.
resubmitHook func(time.Duration, time.Duration) // Method to call upon updating resubmitting interval.
}
func newWorker(config *params.ChainConfig, engine consensus.Engine, eth Backend, mux *event.TypeMux) *worker {
func newWorker(config *params.ChainConfig, engine consensus.Engine, eth Backend, mux *event.TypeMux, recommit time.Duration) *worker {
worker := &worker{
config: config,
engine: engine,
eth: eth,
mux: mux,
chain: eth.BlockChain(),
possibleUncles: make(map[common.Hash]*types.Block),
unconfirmed: newUnconfirmedBlocks(eth.BlockChain(), miningLogAtDepth),
txsCh: make(chan core.NewTxsEvent, txChanSize),
chainHeadCh: make(chan core.ChainHeadEvent, chainHeadChanSize),
chainSideCh: make(chan core.ChainSideEvent, chainSideChanSize),
newWorkCh: make(chan *newWorkReq),
taskCh: make(chan *task),
resultCh: make(chan *task, resultQueueSize),
exitCh: make(chan struct{}),
startCh: make(chan struct{}, 1),
config: config,
engine: engine,
eth: eth,
mux: mux,
chain: eth.BlockChain(),
possibleUncles: make(map[common.Hash]*types.Block),
unconfirmed: newUnconfirmedBlocks(eth.BlockChain(), miningLogAtDepth),
txsCh: make(chan core.NewTxsEvent, txChanSize),
chainHeadCh: make(chan core.ChainHeadEvent, chainHeadChanSize),
chainSideCh: make(chan core.ChainSideEvent, chainSideChanSize),
newWorkCh: make(chan *newWorkReq),
taskCh: make(chan *task),
resultCh: make(chan *task, resultQueueSize),
exitCh: make(chan struct{}),
startCh: make(chan struct{}, 1),
resubmitIntervalCh: make(chan time.Duration),
resubmitAdjustCh: make(chan *intervalAdjust, resubmitAdjustChanSize),
}
// Subscribe NewTxsEvent for tx pool
worker.txsSub = eth.TxPool().SubscribeNewTxsEvent(worker.txsCh)
@@ -163,8 +191,14 @@ func newWorker(config *params.ChainConfig, engine consensus.Engine, eth Backend,
worker.chainHeadSub = eth.BlockChain().SubscribeChainHeadEvent(worker.chainHeadCh)
worker.chainSideSub = eth.BlockChain().SubscribeChainSideEvent(worker.chainSideCh)
// Sanitize recommit interval if the user-specified one is too short.
if recommit < minRecommitInterval {
log.Warn("Sanitizing miner recommit interval", "provided", recommit, "updated", minRecommitInterval)
recommit = minRecommitInterval
}
go worker.mainLoop()
go worker.newWorkLoop()
go worker.newWorkLoop(recommit)
go worker.resultLoop()
go worker.taskLoop()
@@ -188,6 +222,11 @@ func (w *worker) setExtra(extra []byte) {
w.extra = extra
}
// setRecommitInterval updates the interval for miner sealing work recommitting.
func (w *worker) setRecommitInterval(interval time.Duration) {
w.resubmitIntervalCh <- interval
}
// pending returns the pending state and corresponding block.
func (w *worker) pending() (*types.Block, *state.StateDB) {
// return a snapshot to avoid contention on currentMu mutex
@@ -238,35 +277,94 @@ func (w *worker) close() {
}
// newWorkLoop is a standalone goroutine to submit new mining work upon received events.
func (w *worker) newWorkLoop() {
var interrupt *int32
func (w *worker) newWorkLoop(recommit time.Duration) {
var (
interrupt *int32
minRecommit = recommit // minimal resubmit interval specified by user.
)
timer := time.NewTimer(0)
<-timer.C // discard the initial tick
// recommit aborts in-flight transaction execution with given signal and resubmits a new one.
recommit := func(noempty bool, s int32) {
// commit aborts in-flight transaction execution with given signal and resubmits a new one.
commit := func(noempty bool, s int32) {
if interrupt != nil {
atomic.StoreInt32(interrupt, s)
}
interrupt = new(int32)
w.newWorkCh <- &newWorkReq{interrupt: interrupt, noempty: noempty}
timer.Reset(blockRecommitInterval)
timer.Reset(recommit)
atomic.StoreInt32(&w.newTxs, 0)
}
// recalcRecommit recalculates the resubmitting interval upon feedback.
recalcRecommit := func(target float64, inc bool) {
var (
prev = float64(recommit.Nanoseconds())
next float64
)
if inc {
next = prev*(1-intervalAdjustRatio) + intervalAdjustRatio*(target+intervalAdjustBias)
// Recap if interval is larger than the maximum time interval
if next > float64(maxRecommitInterval.Nanoseconds()) {
next = float64(maxRecommitInterval.Nanoseconds())
}
} else {
next = prev*(1-intervalAdjustRatio) + intervalAdjustRatio*(target-intervalAdjustBias)
// Recap if interval is less than the user specified minimum
if next < float64(minRecommit.Nanoseconds()) {
next = float64(minRecommit.Nanoseconds())
}
}
recommit = time.Duration(int64(next))
}
for {
select {
case <-w.startCh:
recommit(false, commitInterruptNewHead)
commit(false, commitInterruptNewHead)
case <-w.chainHeadCh:
recommit(false, commitInterruptNewHead)
commit(false, commitInterruptNewHead)
case <-timer.C:
// If mining is running resubmit a new work cycle periodically to pull in
// higher priced transactions. Disable this overhead for pending blocks.
if w.isRunning() && (w.config.Clique == nil || w.config.Clique.Period > 0) {
recommit(true, commitInterruptResubmit)
// Short circuit if no new transaction arrives.
if atomic.LoadInt32(&w.newTxs) == 0 {
timer.Reset(recommit)
continue
}
commit(true, commitInterruptResubmit)
}
case interval := <-w.resubmitIntervalCh:
// Adjust resubmit interval explicitly by user.
if interval < minRecommitInterval {
log.Warn("Sanitizing miner recommit interval", "provided", interval, "updated", minRecommitInterval)
interval = minRecommitInterval
}
log.Info("Miner recommit interval update", "from", minRecommit, "to", interval)
minRecommit, recommit = interval, interval
if w.resubmitHook != nil {
w.resubmitHook(minRecommit, recommit)
}
case adjust := <-w.resubmitAdjustCh:
// Adjust resubmit interval by feedback.
if adjust.inc {
before := recommit
recalcRecommit(float64(recommit.Nanoseconds())/adjust.ratio, true)
log.Trace("Increase miner recommit interval", "from", before, "to", recommit)
} else {
before := recommit
recalcRecommit(float64(minRecommit.Nanoseconds()), false)
log.Trace("Decrease miner recommit interval", "from", before, "to", recommit)
}
if w.resubmitHook != nil {
w.resubmitHook(minRecommit, recommit)
}
case <-w.exitCh:
@@ -339,6 +437,7 @@ func (w *worker) mainLoop() {
w.commitNewWork(nil, false)
}
}
atomic.AddInt32(&w.newTxs, int32(len(ev.Txs)))
// System stopped
case <-w.exitCh:
@@ -383,7 +482,10 @@ func (w *worker) seal(t *task, stop <-chan struct{}) {
// taskLoop is a standalone goroutine to fetch sealing task from the generator and
// push them to consensus engine.
func (w *worker) taskLoop() {
var stopCh chan struct{}
var (
stopCh chan struct{}
prev common.Hash
)
// interrupt aborts the in-flight sealing task.
interrupt := func() {
@@ -398,8 +500,13 @@ func (w *worker) taskLoop() {
if w.newTaskHook != nil {
w.newTaskHook(task)
}
// Reject duplicate sealing work due to resubmitting.
if task.block.HashNoNonce() == prev {
continue
}
interrupt()
stopCh = make(chan struct{})
prev = task.block.HashNoNonce()
go w.seal(task, stopCh)
case <-w.exitCh:
interrupt()
@@ -414,11 +521,15 @@ func (w *worker) resultLoop() {
for {
select {
case result := <-w.resultCh:
// Short circuit when receiving empty result.
if result == nil {
continue
}
// Short circuit when receiving duplicate result caused by resubmitting.
block := result.block
if w.chain.HasBlock(block.Hash(), block.NumberU64()) {
continue
}
// Update the block hash in all logs since it is now available and not when the
// receipt/log of individual transactions were created.
for _, r := range result.receipts {
@@ -568,8 +679,18 @@ func (w *worker) commitTransactions(txs *types.TransactionsByPriceAndNonce, coin
// (3) worker recreate the mining block with any newly arrived transactions, the interrupt signal is 2.
// For the first two cases, the semi-finished work will be discarded.
// For the third case, the semi-finished work will be submitted to the consensus engine.
// TODO(rjl493456442) give feedback to newWorkLoop to adjust resubmit interval if it is too short.
if interrupt != nil && atomic.LoadInt32(interrupt) != commitInterruptNone {
// Notify resubmit loop to increase resubmitting interval due to too frequent commits.
if atomic.LoadInt32(interrupt) == commitInterruptResubmit {
ratio := float64(w.current.header.GasLimit-w.current.gasPool.Gas()) / float64(w.current.header.GasLimit)
if ratio < 0.1 {
ratio = 0.1
}
w.resubmitAdjustCh <- &intervalAdjust{
ratio: ratio,
inc: true,
}
}
return atomic.LoadInt32(interrupt) == commitInterruptNewHead
}
// If we don't have enough gas for any further transactions then we're done
@@ -644,6 +765,11 @@ func (w *worker) commitTransactions(txs *types.TransactionsByPriceAndNonce, coin
}
go w.mux.Post(core.PendingLogsEvent{Logs: cpy})
}
// Notify resubmit loop to decrease resubmitting interval if current interval is larger
// than the user-specified one.
if interrupt != nil {
w.resubmitAdjustCh <- &intervalAdjust{inc: false}
}
return false
}

106
miner/worker_test.go
View File

@@ -119,7 +119,7 @@ func (b *testWorkerBackend) PostChainEvents(events []interface{}) {
func newTestWorker(t *testing.T, chainConfig *params.ChainConfig, engine consensus.Engine) (*worker, *testWorkerBackend) {
backend := newTestWorkerBackend(t, chainConfig, engine)
backend.txPool.AddLocals(pendingTxs)
w := newWorker(chainConfig, engine, backend, new(event.TypeMux))
w := newWorker(chainConfig, engine, backend, new(event.TypeMux), time.Second)
w.setEtherbase(testBankAddress)
return w, backend
}
@@ -327,7 +327,7 @@ func testRegenerateMiningBlock(t *testing.T, chainConfig *params.ChainConfig, en
}
}
b.txPool.AddLocals(newTxs)
time.Sleep(3 * time.Second)
time.Sleep(time.Second)
select {
case <-taskCh:
@@ -335,3 +335,105 @@ func testRegenerateMiningBlock(t *testing.T, chainConfig *params.ChainConfig, en
t.Error("new task timeout")
}
}
func TestAdjustIntervalEthash(t *testing.T) {
testAdjustInterval(t, ethashChainConfig, ethash.NewFaker())
}
func TestAdjustIntervalClique(t *testing.T) {
testAdjustInterval(t, cliqueChainConfig, clique.New(cliqueChainConfig.Clique, ethdb.NewMemDatabase()))
}
func testAdjustInterval(t *testing.T, chainConfig *params.ChainConfig, engine consensus.Engine) {
defer engine.Close()
w, _ := newTestWorker(t, chainConfig, engine)
defer w.close()
w.skipSealHook = func(task *task) bool {
return true
}
w.fullTaskHook = func() {
time.Sleep(100 * time.Millisecond)
}
var (
progress = make(chan struct{}, 10)
result = make([]float64, 0, 10)
index = 0
start = false
)
w.resubmitHook = func(minInterval time.Duration, recommitInterval time.Duration) {
// Short circuit if interval checking hasn't started.
if !start {
return
}
var wantMinInterval, wantRecommitInterval time.Duration
switch index {
case 0:
wantMinInterval, wantRecommitInterval = 3*time.Second, 3*time.Second
case 1:
origin := float64(3 * time.Second.Nanoseconds())
estimate := origin*(1-intervalAdjustRatio) + intervalAdjustRatio*(origin/0.8+intervalAdjustBias)
wantMinInterval, wantRecommitInterval = 3*time.Second, time.Duration(int(estimate))*time.Nanosecond
case 2:
estimate := result[index-1]
min := float64(3 * time.Second.Nanoseconds())
estimate = estimate*(1-intervalAdjustRatio) + intervalAdjustRatio*(min-intervalAdjustBias)
wantMinInterval, wantRecommitInterval = 3*time.Second, time.Duration(int(estimate))*time.Nanosecond
case 3:
wantMinInterval, wantRecommitInterval = time.Second, time.Second
}
// Check interval
if minInterval != wantMinInterval {
t.Errorf("resubmit min interval mismatch want %s has %s", wantMinInterval, minInterval)
}
if recommitInterval != wantRecommitInterval {
t.Errorf("resubmit interval mismatch want %s has %s", wantRecommitInterval, recommitInterval)
}
result = append(result, float64(recommitInterval.Nanoseconds()))
index += 1
progress <- struct{}{}
}
// Ensure worker has finished initialization
for {
b := w.pendingBlock()
if b != nil && b.NumberU64() == 1 {
break
}
}
w.start()
time.Sleep(time.Second)
start = true
w.setRecommitInterval(3 * time.Second)
select {
case <-progress:
case <-time.NewTimer(time.Second).C:
t.Error("interval reset timeout")
}
w.resubmitAdjustCh <- &intervalAdjust{inc: true, ratio: 0.8}
select {
case <-progress:
case <-time.NewTimer(time.Second).C:
t.Error("interval reset timeout")
}
w.resubmitAdjustCh <- &intervalAdjust{inc: false}
select {
case <-progress:
case <-time.NewTimer(time.Second).C:
t.Error("interval reset timeout")
}
w.setRecommitInterval(500 * time.Millisecond)
select {
case <-progress:
case <-time.NewTimer(time.Second).C:
t.Error("interval reset timeout")
}
}