You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

313 lines
7.9 KiB

8 years ago
8 years ago
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
8 years ago
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
8 years ago
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
8 years ago
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
8 years ago
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
8 years ago
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
8 years ago
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
8 years ago
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
8 years ago
8 years ago
8 years ago
8 years ago
8 years ago
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
8 years ago
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
8 years ago
  1. package consensus
  2. import (
  3. "bytes"
  4. "encoding/binary"
  5. "fmt"
  6. "hash/crc32"
  7. "io"
  8. "path/filepath"
  9. "time"
  10. "github.com/pkg/errors"
  11. wire "github.com/tendermint/go-wire"
  12. "github.com/tendermint/tendermint/types"
  13. auto "github.com/tendermint/tmlibs/autofile"
  14. cmn "github.com/tendermint/tmlibs/common"
  15. )
  16. //--------------------------------------------------------
  17. // types and functions for savings consensus messages
  18. var (
  19. walSeparator = []byte{55, 127, 6, 130} // 0x377f0682 - magic number
  20. )
  21. type TimedWALMessage struct {
  22. Time time.Time `json:"time"` // for debugging purposes
  23. Msg WALMessage `json:"msg"`
  24. }
  25. // EndHeightMessage marks the end of the given height inside WAL.
  26. // @internal used by scripts/cutWALUntil util.
  27. type EndHeightMessage struct {
  28. Height uint64 `json:"height"`
  29. }
  30. type WALMessage interface{}
  31. var _ = wire.RegisterInterface(
  32. struct{ WALMessage }{},
  33. wire.ConcreteType{types.EventDataRoundState{}, 0x01},
  34. wire.ConcreteType{msgInfo{}, 0x02},
  35. wire.ConcreteType{timeoutInfo{}, 0x03},
  36. wire.ConcreteType{EndHeightMessage{}, 0x04},
  37. )
  38. //--------------------------------------------------------
  39. // Simple write-ahead logger
  40. // WAL is an interface for any write-ahead logger.
  41. type WAL interface {
  42. Save(WALMessage)
  43. Group() *auto.Group
  44. SearchForEndHeight(height uint64) (gr *auto.GroupReader, found bool, err error)
  45. Start() (bool, error)
  46. Stop() bool
  47. Wait()
  48. }
  49. // Write ahead logger writes msgs to disk before they are processed.
  50. // Can be used for crash-recovery and deterministic replay
  51. // TODO: currently the wal is overwritten during replay catchup
  52. // give it a mode so it's either reading or appending - must read to end to start appending again
  53. type baseWAL struct {
  54. cmn.BaseService
  55. group *auto.Group
  56. light bool // ignore block parts
  57. enc *WALEncoder
  58. }
  59. func NewWAL(walFile string, light bool) (*baseWAL, error) {
  60. err := cmn.EnsureDir(filepath.Dir(walFile), 0700)
  61. if err != nil {
  62. return nil, errors.Wrap(err, "failed to ensure WAL directory is in place")
  63. }
  64. group, err := auto.OpenGroup(walFile)
  65. if err != nil {
  66. return nil, err
  67. }
  68. wal := &baseWAL{
  69. group: group,
  70. light: light,
  71. enc: NewWALEncoder(group),
  72. }
  73. wal.BaseService = *cmn.NewBaseService(nil, "baseWAL", wal)
  74. return wal, nil
  75. }
  76. func (wal *baseWAL) Group() *auto.Group {
  77. return wal.group
  78. }
  79. func (wal *baseWAL) OnStart() error {
  80. size, err := wal.group.Head.Size()
  81. if err != nil {
  82. return err
  83. } else if size == 0 {
  84. wal.Save(EndHeightMessage{0})
  85. }
  86. _, err = wal.group.Start()
  87. return err
  88. }
  89. func (wal *baseWAL) OnStop() {
  90. wal.BaseService.OnStop()
  91. wal.group.Stop()
  92. }
  93. // called in newStep and for each pass in receiveRoutine
  94. func (wal *baseWAL) Save(msg WALMessage) {
  95. if wal == nil {
  96. return
  97. }
  98. if wal.light {
  99. // in light mode we only write new steps, timeouts, and our own votes (no proposals, block parts)
  100. if mi, ok := msg.(msgInfo); ok {
  101. if mi.PeerKey != "" {
  102. return
  103. }
  104. }
  105. }
  106. // Write the wal message
  107. if err := wal.enc.Encode(&TimedWALMessage{time.Now(), msg}); err != nil {
  108. cmn.PanicQ(cmn.Fmt("Error writing msg to consensus wal: %v \n\nMessage: %v", err, msg))
  109. }
  110. // TODO: only flush when necessary
  111. if err := wal.group.Flush(); err != nil {
  112. cmn.PanicQ(cmn.Fmt("Error flushing consensus wal buf to file. Error: %v \n", err))
  113. }
  114. }
  115. // SearchForEndHeight searches for the EndHeightMessage with the height and
  116. // returns an auto.GroupReader, whenever it was found or not and an error.
  117. // Group reader will be nil if found equals false.
  118. //
  119. // CONTRACT: caller must close group reader.
  120. func (wal *baseWAL) SearchForEndHeight(height uint64) (gr *auto.GroupReader, found bool, err error) {
  121. var msg *TimedWALMessage
  122. // NOTE: starting from the last file in the group because we're usually
  123. // searching for the last height. See replay.go
  124. min, max := wal.group.MinIndex(), wal.group.MaxIndex()
  125. wal.Logger.Debug("Searching for height", "height", height, "min", min, "max", max)
  126. for index := max; index >= min; index-- {
  127. gr, err = wal.group.NewReader(index)
  128. if err != nil {
  129. return nil, false, err
  130. }
  131. dec := NewWALDecoder(gr)
  132. for {
  133. msg, err = dec.Decode()
  134. if err == io.EOF {
  135. // check next file
  136. break
  137. }
  138. if err != nil {
  139. gr.Close()
  140. return nil, false, err
  141. }
  142. if m, ok := msg.Msg.(EndHeightMessage); ok {
  143. if m.Height == height { // found
  144. wal.Logger.Debug("Found", "height", height, "index", index)
  145. return gr, true, nil
  146. }
  147. }
  148. }
  149. gr.Close()
  150. }
  151. return nil, false, nil
  152. }
  153. ///////////////////////////////////////////////////////////////////////////////
  154. // A WALEncoder writes custom-encoded WAL messages to an output stream.
  155. //
  156. // Format: 4 bytes CRC sum + 4 bytes length + arbitrary-length value (go-wire encoded)
  157. type WALEncoder struct {
  158. wr io.Writer
  159. }
  160. // NewWALEncoder returns a new encoder that writes to wr.
  161. func NewWALEncoder(wr io.Writer) *WALEncoder {
  162. return &WALEncoder{wr}
  163. }
  164. // Encode writes the custom encoding of v to the stream.
  165. func (enc *WALEncoder) Encode(v interface{}) error {
  166. data := wire.BinaryBytes(v)
  167. crc := crc32.Checksum(data, crc32c)
  168. length := uint32(len(data))
  169. totalLength := 8 + int(length)
  170. msg := make([]byte, totalLength)
  171. binary.BigEndian.PutUint32(msg[0:4], crc)
  172. binary.BigEndian.PutUint32(msg[4:8], length)
  173. copy(msg[8:], data)
  174. _, err := enc.wr.Write(msg)
  175. if err == nil {
  176. // TODO [Anton Kaliaev 23 Oct 2017]: remove separator
  177. _, err = enc.wr.Write(walSeparator)
  178. }
  179. return err
  180. }
  181. ///////////////////////////////////////////////////////////////////////////////
  182. // A WALDecoder reads and decodes custom-encoded WAL messages from an input
  183. // stream. See WALEncoder for the format used.
  184. //
  185. // It will also compare the checksums and make sure data size is equal to the
  186. // length from the header. If that is not the case, error will be returned.
  187. type WALDecoder struct {
  188. rd io.Reader
  189. }
  190. // NewWALDecoder returns a new decoder that reads from rd.
  191. func NewWALDecoder(rd io.Reader) *WALDecoder {
  192. return &WALDecoder{rd}
  193. }
  194. // Decode reads the next custom-encoded value from its reader and returns it.
  195. func (dec *WALDecoder) Decode() (*TimedWALMessage, error) {
  196. b := make([]byte, 4)
  197. n, err := dec.rd.Read(b)
  198. if err == io.EOF {
  199. return nil, err
  200. }
  201. if err != nil {
  202. return nil, fmt.Errorf("failed to read checksum: %v", err)
  203. }
  204. crc := binary.BigEndian.Uint32(b)
  205. b = make([]byte, 4)
  206. n, err = dec.rd.Read(b)
  207. if err == io.EOF {
  208. return nil, err
  209. }
  210. if err != nil {
  211. return nil, fmt.Errorf("failed to read length: %v", err)
  212. }
  213. length := binary.BigEndian.Uint32(b)
  214. data := make([]byte, length)
  215. n, err = dec.rd.Read(data)
  216. if err == io.EOF {
  217. return nil, err
  218. }
  219. if err != nil {
  220. return nil, fmt.Errorf("not enough bytes for data: %v (want: %d, read: %v)", err, length, n)
  221. }
  222. // check checksum before decoding data
  223. actualCRC := crc32.Checksum(data, crc32c)
  224. if actualCRC != crc {
  225. return nil, fmt.Errorf("checksums do not match: (read: %v, actual: %v)", crc, actualCRC)
  226. }
  227. var nn int
  228. var res *TimedWALMessage
  229. res = wire.ReadBinary(&TimedWALMessage{}, bytes.NewBuffer(data), int(length), &nn, &err).(*TimedWALMessage)
  230. if err != nil {
  231. return nil, fmt.Errorf("failed to decode data: %v", err)
  232. }
  233. // TODO [Anton Kaliaev 23 Oct 2017]: remove separator
  234. if err = readSeparator(dec.rd); err != nil {
  235. return nil, err
  236. }
  237. return res, err
  238. }
  239. // readSeparator reads a separator from r. It returns any error from underlying
  240. // reader or if it's not a separator.
  241. func readSeparator(r io.Reader) error {
  242. b := make([]byte, len(walSeparator))
  243. _, err := r.Read(b)
  244. if err != nil {
  245. return fmt.Errorf("failed to read separator: %v", err)
  246. }
  247. if !bytes.Equal(b, walSeparator) {
  248. return fmt.Errorf("not a separator: %v", b)
  249. }
  250. return nil
  251. }
  252. type nilWAL struct{}
  253. func (nilWAL) Save(m WALMessage) {}
  254. func (nilWAL) Group() *auto.Group { return nil }
  255. func (nilWAL) SearchForEndHeight(height uint64) (gr *auto.GroupReader, found bool, err error) {
  256. return nil, false, nil
  257. }
  258. func (nilWAL) Start() (bool, error) { return true, nil }
  259. func (nilWAL) Stop() bool { return true }
  260. func (nilWAL) Wait() {}