tendermint/state/execution.go

315 lines
9.6 KiB
Go
Raw Normal View History

package state
import (
"errors"
"fmt"
fail "github.com/ebuchman/fail-test"
2017-02-14 12:33:14 -08:00
abci "github.com/tendermint/abci/types"
crypto "github.com/tendermint/go-crypto"
2015-12-01 20:12:01 -08:00
"github.com/tendermint/tendermint/proxy"
2015-04-01 17:30:16 -07:00
"github.com/tendermint/tendermint/types"
cmn "github.com/tendermint/tmlibs/common"
2017-05-02 00:53:32 -07:00
"github.com/tendermint/tmlibs/log"
)
//--------------------------------------------------
// Execute the block
// ValExecBlock executes the block, but does NOT mutate State.
// + validates the block
// + executes block.Txs on the proxyAppConn
new pubsub package comment out failing consensus tests for now rewrite rpc httpclient to use new pubsub package import pubsub as tmpubsub, query as tmquery make event IDs constants EventKey -> EventTypeKey rename EventsPubsub to PubSub mempool does not use pubsub rename eventsSub to pubsub new subscribe API fix channel size issues and consensus tests bugs refactor rpc client add missing discardFromChan method add mutex rename pubsub to eventBus remove IsRunning from WSRPCConnection interface (not needed) add a comment in broadcastNewRoundStepsAndVotes rename registerEventCallbacks to broadcastNewRoundStepsAndVotes See https://dave.cheney.net/2014/03/19/channel-axioms stop eventBuses after reactor tests remove unnecessary Unsubscribe return subscribe helper function move discardFromChan to where it is used subscribe now returns an err this gives us ability to refuse to subscribe if pubsub is at its max capacity. use context for control overflow cache queries handle err when subscribing in replay_test rename testClientID to testSubscriber extract var set channel buffer capacity to 1 in replay_file fix byzantine_test unsubscribe from single event, not all events refactor httpclient to return events to appropriate channels return failing testReplayCrashBeforeWriteVote test fix TestValidatorSetChanges refactor code a bit fix testReplayCrashBeforeWriteVote add comment fix TestValidatorSetChanges fixes from Bucky's review update comment [ci skip] test TxEventBuffer update changelog fix TestValidatorSetChanges (2nd attempt) only do wg.Done when no errors benchmark event bus create pubsub server inside NewEventBus only expose config params (later if needed) set buffer capacity to 0 so we are not testing cache new tx event format: key = "Tx" plus a tag {"tx.hash": XYZ} This should allow to subscribe to all transactions! or a specific one using a query: "tm.events.type = Tx and tx.hash = '013ABF99434...'" use TimeoutCommit instead of afterPublishEventNewBlockTimeout TimeoutCommit is the time a node waits after committing a block, before it goes into the next height. So it will finish everything from the last block, but then wait a bit. The idea is this gives it time to hear more votes from other validators, to strengthen the commit it includes in the next block. But it also gives it time to hear about new transactions. waitForBlockWithUpdatedVals rewrite WAL crash tests Task: test that we can recover from any WAL crash. Solution: the old tests were relying on event hub being run in the same thread (we were injecting the private validator's last signature). when considering a rewrite, we considered two possible solutions: write a "fuzzy" testing system where WAL is crashing upon receiving a new message, or inject failures and trigger them in tests using something like https://github.com/coreos/gofail. remove sleep no cs.Lock around wal.Save test different cases (empty block, non-empty block, ...) comments add comments test 4 cases: empty block, non-empty block, non-empty block with smaller part size, many blocks fixes as per Bucky's last review reset subscriptions on UnsubscribeAll use a simple counter to track message for which we panicked also, set a smaller part size for all test cases
2017-06-26 08:00:30 -07:00
func (s *State) ValExecBlock(txEventPublisher types.TxEventPublisher, proxyAppConn proxy.AppConnConsensus, block *types.Block) (*ABCIResponses, error) {
2015-12-01 20:12:01 -08:00
// Validate the block.
2016-11-19 16:32:35 -08:00
if err := s.validateBlock(block); err != nil {
return nil, ErrInvalidBlock(err)
}
2015-12-01 20:12:01 -08:00
// Execute the block txs
abciResponses, err := execBlockOnProxyApp(txEventPublisher, proxyAppConn, block, s.logger, s.LastValidators)
2015-12-01 20:12:01 -08:00
if err != nil {
// There was some error in proxyApp
// TODO Report error and wait for proxyApp to be available.
return nil, ErrProxyAppConn(err)
2015-12-01 20:12:01 -08:00
}
return abciResponses, nil
}
// Executes block's transactions on proxyAppConn.
// Returns a list of transaction results and updates to the validator set
// TODO: Generate a bitmap or otherwise store tx validity in state.
func execBlockOnProxyApp(txEventPublisher types.TxEventPublisher, proxyAppConn proxy.AppConnConsensus, block *types.Block, logger log.Logger, lastValidators *types.ValidatorSet) (*ABCIResponses, error) {
var validTxs, invalidTxs = 0, 0
2015-12-01 20:12:01 -08:00
txIndex := 0
abciResponses := NewABCIResponses(block)
2015-12-01 20:12:01 -08:00
// Execute transactions and get hash
2017-01-12 12:53:32 -08:00
proxyCb := func(req *abci.Request, res *abci.Response) {
2016-05-14 09:33:27 -07:00
switch r := res.Value.(type) {
2017-01-12 12:55:03 -08:00
case *abci.Response_DeliverTx:
2016-01-25 14:34:08 -08:00
// TODO: make use of res.Log
// TODO: make use of this info
// Blocks may include invalid txs.
2017-01-12 12:55:03 -08:00
// reqDeliverTx := req.(abci.RequestDeliverTx)
txResult := r.DeliverTx
2017-12-01 22:47:55 -08:00
if txResult.Code == abci.CodeTypeOK {
validTxs++
} else {
2017-05-02 00:53:32 -07:00
logger.Debug("Invalid tx", "code", txResult.Code, "log", txResult.Log)
invalidTxs++
}
2016-06-27 17:43:09 -07:00
// NOTE: if we count we can access the tx from the block instead of
// pulling it from the req
txEventPublisher.PublishEventTx(types.EventDataTx{types.TxResult{
Height: block.Height,
Index: uint32(txIndex),
Tx: types.Tx(req.GetDeliverTx().Tx),
Result: *txResult,
}})
2017-11-15 12:10:54 -08:00
abciResponses.DeliverTx[txIndex] = txResult
txIndex++
2015-12-01 20:12:01 -08:00
}
}
proxyAppConn.SetResponseCallback(proxyCb)
// determine validators who did not sign last block
absentVals := make([]int32, 0)
for valA, _ := range lastValidators.Validators {
found := false
for _, voteB := range block.LastCommit.Precommits {
valB := voteB.ValidatorIndex
if valA == valB {
found = true
break
}
}
if !found {
absentVals = append(absentVals, int32(valA))
}
}
2016-11-03 16:51:22 -07:00
// Begin block
2017-11-29 09:22:52 -08:00
_, err := proxyAppConn.BeginBlockSync(abci.RequestBeginBlock{
Hash: block.Hash(),
Header: types.TM2PB.Header(block.Header),
AbsentValidators: absentVals,
ByzantineValidators: nil,
})
2016-11-03 16:51:22 -07:00
if err != nil {
logger.Error("Error in proxyAppConn.BeginBlock", "err", err)
return nil, err
2016-11-03 16:51:22 -07:00
}
2016-03-05 20:57:36 -08:00
// Run txs of block
for _, tx := range block.Txs {
2017-01-12 12:55:03 -08:00
proxyAppConn.DeliverTxAsync(tx)
if err := proxyAppConn.Error(); err != nil {
return nil, err
2015-12-01 20:12:01 -08:00
}
}
2016-03-05 20:57:36 -08:00
// End block
2017-12-01 22:47:55 -08:00
abciResponses.EndBlock, err = proxyAppConn.EndBlockSync(abci.RequestEndBlock{block.Height})
2016-03-05 20:57:36 -08:00
if err != nil {
logger.Error("Error in proxyAppConn.EndBlock", "err", err)
return nil, err
2016-03-05 20:57:36 -08:00
}
2016-09-11 12:32:33 -07:00
valDiff := abciResponses.EndBlock.Diffs
2017-06-13 23:41:36 -07:00
logger.Info("Executed block", "height", block.Height, "validTxs", validTxs, "invalidTxs", invalidTxs)
if len(valDiff) > 0 {
2017-05-02 00:53:32 -07:00
logger.Info("Update to validator set", "updates", abci.ValidatorsString(valDiff))
2016-11-22 15:55:42 -08:00
}
return abciResponses, nil
2016-11-19 16:32:35 -08:00
}
2017-01-12 12:53:32 -08:00
func updateValidators(validators *types.ValidatorSet, changedValidators []*abci.Validator) error {
2016-11-19 16:32:35 -08:00
// TODO: prevent change of 1/3+ at once
for _, v := range changedValidators {
pubkey, err := crypto.PubKeyFromBytes(v.PubKey) // NOTE: expects go-wire encoded pubkey
if err != nil {
return err
2016-11-19 16:32:35 -08:00
}
address := pubkey.Address()
power := int64(v.Power)
// mind the overflow from int64
2016-11-19 16:32:35 -08:00
if power < 0 {
return errors.New(cmn.Fmt("Power (%d) overflows int64", v.Power))
2016-11-19 16:32:35 -08:00
}
_, val := validators.GetByAddress(address)
if val == nil {
// add val
added := validators.Add(types.NewValidator(pubkey, power))
if !added {
return errors.New(cmn.Fmt("Failed to add new validator %X with voting power %d", address, power))
2016-11-19 16:32:35 -08:00
}
} else if v.Power == 0 {
// remove val
_, removed := validators.Remove(address)
if !removed {
return errors.New(cmn.Fmt("Failed to remove validator %X)"))
2016-11-19 16:32:35 -08:00
}
} else {
// update val
val.VotingPower = power
updated := validators.Update(val)
if !updated {
return errors.New(cmn.Fmt("Failed to update validator %X with voting power %d", address, power))
2016-11-19 16:32:35 -08:00
}
}
}
return nil
2015-12-01 20:12:01 -08:00
}
2017-09-21 05:49:19 -07:00
// return a bit array of validators that signed the last commit
// NOTE: assumes commits have already been authenticated
/* function is currently unused
2017-09-21 05:49:19 -07:00
func commitBitArrayFromBlock(block *types.Block) *cmn.BitArray {
signed := cmn.NewBitArray(len(block.LastCommit.Precommits))
for i, precommit := range block.LastCommit.Precommits {
if precommit != nil {
signed.SetIndex(i, true) // val_.LastCommitHeight = block.Height - 1
}
}
return signed
}
*/
2017-09-21 05:49:19 -07:00
//-----------------------------------------------------
// Validate block
2017-08-21 13:12:07 -07:00
// ValidateBlock validates the block against the state.
func (s *State) ValidateBlock(block *types.Block) error {
return s.validateBlock(block)
}
func (s *State) validateBlock(block *types.Block) error {
// Basic block validation.
err := block.ValidateBasic(s.ChainID, s.LastBlockHeight,
s.LastBlockTotalTx, s.LastBlockID, s.LastBlockTime, s.AppHash)
if err != nil {
return err
}
// Validate block LastCommit.
if block.Height == 1 {
if len(block.LastCommit.Precommits) != 0 {
return errors.New("Block at height 1 (first block) should have no LastCommit precommits")
}
} else {
if len(block.LastCommit.Precommits) != s.LastValidators.Size() {
return errors.New(cmn.Fmt("Invalid block commit size. Expected %v, got %v",
s.LastValidators.Size(), len(block.LastCommit.Precommits)))
}
err := s.LastValidators.VerifyCommit(
s.ChainID, s.LastBlockID, block.Height-1, block.LastCommit)
if err != nil {
return err
}
}
return nil
2015-12-01 20:12:01 -08:00
}
2016-08-23 18:44:07 -07:00
//-----------------------------------------------------------------------------
// ApplyBlock validates & executes the block, updates state w/ ABCI responses,
// then commits and updates the mempool atomically, then saves state.
2017-08-21 13:12:07 -07:00
// ApplyBlock validates the block against the state, executes it against the app,
// commits it, and saves the block and state. It's the only function that needs to be called
// from outside this package to process and commit an entire block.
new pubsub package comment out failing consensus tests for now rewrite rpc httpclient to use new pubsub package import pubsub as tmpubsub, query as tmquery make event IDs constants EventKey -> EventTypeKey rename EventsPubsub to PubSub mempool does not use pubsub rename eventsSub to pubsub new subscribe API fix channel size issues and consensus tests bugs refactor rpc client add missing discardFromChan method add mutex rename pubsub to eventBus remove IsRunning from WSRPCConnection interface (not needed) add a comment in broadcastNewRoundStepsAndVotes rename registerEventCallbacks to broadcastNewRoundStepsAndVotes See https://dave.cheney.net/2014/03/19/channel-axioms stop eventBuses after reactor tests remove unnecessary Unsubscribe return subscribe helper function move discardFromChan to where it is used subscribe now returns an err this gives us ability to refuse to subscribe if pubsub is at its max capacity. use context for control overflow cache queries handle err when subscribing in replay_test rename testClientID to testSubscriber extract var set channel buffer capacity to 1 in replay_file fix byzantine_test unsubscribe from single event, not all events refactor httpclient to return events to appropriate channels return failing testReplayCrashBeforeWriteVote test fix TestValidatorSetChanges refactor code a bit fix testReplayCrashBeforeWriteVote add comment fix TestValidatorSetChanges fixes from Bucky's review update comment [ci skip] test TxEventBuffer update changelog fix TestValidatorSetChanges (2nd attempt) only do wg.Done when no errors benchmark event bus create pubsub server inside NewEventBus only expose config params (later if needed) set buffer capacity to 0 so we are not testing cache new tx event format: key = "Tx" plus a tag {"tx.hash": XYZ} This should allow to subscribe to all transactions! or a specific one using a query: "tm.events.type = Tx and tx.hash = '013ABF99434...'" use TimeoutCommit instead of afterPublishEventNewBlockTimeout TimeoutCommit is the time a node waits after committing a block, before it goes into the next height. So it will finish everything from the last block, but then wait a bit. The idea is this gives it time to hear more votes from other validators, to strengthen the commit it includes in the next block. But it also gives it time to hear about new transactions. waitForBlockWithUpdatedVals rewrite WAL crash tests Task: test that we can recover from any WAL crash. Solution: the old tests were relying on event hub being run in the same thread (we were injecting the private validator's last signature). when considering a rewrite, we considered two possible solutions: write a "fuzzy" testing system where WAL is crashing upon receiving a new message, or inject failures and trigger them in tests using something like https://github.com/coreos/gofail. remove sleep no cs.Lock around wal.Save test different cases (empty block, non-empty block, ...) comments add comments test 4 cases: empty block, non-empty block, non-empty block with smaller part size, many blocks fixes as per Bucky's last review reset subscriptions on UnsubscribeAll use a simple counter to track message for which we panicked also, set a smaller part size for all test cases
2017-06-26 08:00:30 -07:00
func (s *State) ApplyBlock(txEventPublisher types.TxEventPublisher, proxyAppConn proxy.AppConnConsensus,
block *types.Block, partsHeader types.PartSetHeader, mempool types.Mempool) error {
new pubsub package comment out failing consensus tests for now rewrite rpc httpclient to use new pubsub package import pubsub as tmpubsub, query as tmquery make event IDs constants EventKey -> EventTypeKey rename EventsPubsub to PubSub mempool does not use pubsub rename eventsSub to pubsub new subscribe API fix channel size issues and consensus tests bugs refactor rpc client add missing discardFromChan method add mutex rename pubsub to eventBus remove IsRunning from WSRPCConnection interface (not needed) add a comment in broadcastNewRoundStepsAndVotes rename registerEventCallbacks to broadcastNewRoundStepsAndVotes See https://dave.cheney.net/2014/03/19/channel-axioms stop eventBuses after reactor tests remove unnecessary Unsubscribe return subscribe helper function move discardFromChan to where it is used subscribe now returns an err this gives us ability to refuse to subscribe if pubsub is at its max capacity. use context for control overflow cache queries handle err when subscribing in replay_test rename testClientID to testSubscriber extract var set channel buffer capacity to 1 in replay_file fix byzantine_test unsubscribe from single event, not all events refactor httpclient to return events to appropriate channels return failing testReplayCrashBeforeWriteVote test fix TestValidatorSetChanges refactor code a bit fix testReplayCrashBeforeWriteVote add comment fix TestValidatorSetChanges fixes from Bucky's review update comment [ci skip] test TxEventBuffer update changelog fix TestValidatorSetChanges (2nd attempt) only do wg.Done when no errors benchmark event bus create pubsub server inside NewEventBus only expose config params (later if needed) set buffer capacity to 0 so we are not testing cache new tx event format: key = "Tx" plus a tag {"tx.hash": XYZ} This should allow to subscribe to all transactions! or a specific one using a query: "tm.events.type = Tx and tx.hash = '013ABF99434...'" use TimeoutCommit instead of afterPublishEventNewBlockTimeout TimeoutCommit is the time a node waits after committing a block, before it goes into the next height. So it will finish everything from the last block, but then wait a bit. The idea is this gives it time to hear more votes from other validators, to strengthen the commit it includes in the next block. But it also gives it time to hear about new transactions. waitForBlockWithUpdatedVals rewrite WAL crash tests Task: test that we can recover from any WAL crash. Solution: the old tests were relying on event hub being run in the same thread (we were injecting the private validator's last signature). when considering a rewrite, we considered two possible solutions: write a "fuzzy" testing system where WAL is crashing upon receiving a new message, or inject failures and trigger them in tests using something like https://github.com/coreos/gofail. remove sleep no cs.Lock around wal.Save test different cases (empty block, non-empty block, ...) comments add comments test 4 cases: empty block, non-empty block, non-empty block with smaller part size, many blocks fixes as per Bucky's last review reset subscriptions on UnsubscribeAll use a simple counter to track message for which we panicked also, set a smaller part size for all test cases
2017-06-26 08:00:30 -07:00
abciResponses, err := s.ValExecBlock(txEventPublisher, proxyAppConn, block)
if err != nil {
return fmt.Errorf("Exec failed for application: %v", err)
}
fail.Fail() // XXX
// save the results before we commit
s.SaveABCIResponses(abciResponses)
fail.Fail() // XXX
// now update the block and validators
s.SetBlockAndValidators(block.Header, partsHeader, abciResponses)
// lock mempool, commit state, update mempoool
err = s.CommitStateUpdateMempool(proxyAppConn, block, mempool)
if err != nil {
return fmt.Errorf("Commit failed for application: %v", err)
}
fail.Fail() // XXX
2017-09-04 15:27:04 -07:00
// save the state and the validators
s.Save()
return nil
}
2016-08-23 18:44:07 -07:00
2017-08-21 13:12:07 -07:00
// CommitStateUpdateMempool locks the mempool, runs the ABCI Commit message, and updates the mempool.
// The Mempool must be locked during commit and update because state is typically reset on Commit and old txs must be replayed
// against committed state before new txs are run in the mempool, lest they be invalid.
func (s *State) CommitStateUpdateMempool(proxyAppConn proxy.AppConnConsensus, block *types.Block, mempool types.Mempool) error {
2016-08-24 21:18:03 -07:00
mempool.Lock()
defer mempool.Unlock()
// Commit block, get hash back
res, err := proxyAppConn.CommitSync()
if err != nil {
s.logger.Error("Client error during proxyAppConn.CommitSync", "err", err)
return err
}
2016-08-24 21:18:03 -07:00
if res.IsErr() {
s.logger.Error("Error in proxyAppConn.CommitSync", "err", res)
2016-08-24 21:18:03 -07:00
return res
}
if res.Log != "" {
2017-05-02 00:53:32 -07:00
s.logger.Debug("Commit.Log: " + res.Log)
2016-08-24 21:18:03 -07:00
}
2017-05-26 08:57:41 -07:00
s.logger.Info("Committed state", "height", block.Height, "txs", block.NumTxs, "hash", res.Data)
2016-08-24 21:18:03 -07:00
// Set the state's new AppHash
s.AppHash = res.Data
// Update mempool.
2017-09-21 08:42:44 -07:00
return mempool.Update(block.Height, block.Txs)
2016-08-24 21:18:03 -07:00
}
2017-08-21 13:12:07 -07:00
// ExecCommitBlock executes and commits a block on the proxyApp without validating or mutating the state.
// It returns the application root hash (result of abci.Commit).
func ExecCommitBlock(appConnConsensus proxy.AppConnConsensus, block *types.Block, logger log.Logger, lastValidators *types.ValidatorSet) ([]byte, error) {
_, err := execBlockOnProxyApp(types.NopEventBus{}, appConnConsensus, block, logger, lastValidators)
if err != nil {
2017-05-02 00:53:32 -07:00
logger.Error("Error executing block on proxy app", "height", block.Height, "err", err)
return nil, err
}
// Commit block, get hash back
res, err := appConnConsensus.CommitSync()
if err != nil {
logger.Error("Client error during proxyAppConn.CommitSync", "err", res)
return nil, err
}
if res.IsErr() {
logger.Error("Error in proxyAppConn.CommitSync", "err", res)
return nil, res
}
if res.Log != "" {
2017-05-02 00:53:32 -07:00
logger.Info("Commit.Log: " + res.Log)
}
return res.Data, nil
}