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.

339 lines
9.3 KiB

7 years ago
7 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
7 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
7 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
7 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
7 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
7 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
7 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
7 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
7 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
7 years ago
fix non deterministic test failures and race in privval socket (#3258) * node: decrease retry conn timeout in test Should fix #3256 The retry timeout was set to the default, which is the same as the accept timeout, so it's no wonder this would fail. Here we decrease the retry timeout so we can try many times before the accept timeout. * p2p: increase handshake timeout in test This fails sometimes, presumably because the handshake timeout is so low (only 50ms). So increase it to 1s. Should fix #3187 * privval: fix race with ping. closes #3237 Pings happen in a go-routine and can happen concurrently with other messages. Since we use a request/response protocol, we expect to send a request and get back the corresponding response. But with pings happening concurrently, this assumption could be violated. We were using a mutex, but only a RWMutex, where the RLock was being held for sending messages - this was to allow the underlying connection to be replaced if it fails. Turns out we actually need to use a full lock (not just a read lock) to prevent multiple requests from happening concurrently. * node: fix test name. DelayedStop -> DelayedStart * autofile: Wait() method In the TestWALTruncate in consensus/wal_test.go we remove the WAL directory at the end of the test. However the wal.Stop() does not properly wait for the autofile group to finish shutting down. Hence it was possible that the group's go-routine is still running when the cleanup happens, which causes a panic since the directory disappeared. Here we add a Wait() method to properly wait until the go-routine exits so we can safely clean up. This fixes #2852.
6 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
7 years ago
fix non deterministic test failures and race in privval socket (#3258) * node: decrease retry conn timeout in test Should fix #3256 The retry timeout was set to the default, which is the same as the accept timeout, so it's no wonder this would fail. Here we decrease the retry timeout so we can try many times before the accept timeout. * p2p: increase handshake timeout in test This fails sometimes, presumably because the handshake timeout is so low (only 50ms). So increase it to 1s. Should fix #3187 * privval: fix race with ping. closes #3237 Pings happen in a go-routine and can happen concurrently with other messages. Since we use a request/response protocol, we expect to send a request and get back the corresponding response. But with pings happening concurrently, this assumption could be violated. We were using a mutex, but only a RWMutex, where the RLock was being held for sending messages - this was to allow the underlying connection to be replaced if it fails. Turns out we actually need to use a full lock (not just a read lock) to prevent multiple requests from happening concurrently. * node: fix test name. DelayedStop -> DelayedStart * autofile: Wait() method In the TestWALTruncate in consensus/wal_test.go we remove the WAL directory at the end of the test. However the wal.Stop() does not properly wait for the autofile group to finish shutting down. Hence it was possible that the group's go-routine is still running when the cleanup happens, which causes a panic since the directory disappeared. Here we add a Wait() method to properly wait until the go-routine exits so we can safely clean up. This fixes #2852.
6 years ago
6 years ago
6 years ago
6 years ago
6 years ago
8 years ago
6 years ago
WAL: better errors and new fail point (#3246) * privval: more info in errors * wal: change Debug logs to Info * wal: log and return error on corrupted wal instead of panicing * fail: Exit right away instead of sending interupt * consensus: FAIL before handling our own vote allows to replicate #3089: - run using `FAIL_TEST_INDEX=0` - delete some bytes from the end of the WAL - start normally Results in logs like: ``` I[2019-02-03|18:12:58.225] Searching for height module=consensus wal=/Users/ethanbuchman/.tendermint/data/cs.wal/wal height=1 min=0 max=0 E[2019-02-03|18:12:58.225] Error on catchup replay. Proceeding to start ConsensusState anyway module=consensus err="failed to read data: EOF" I[2019-02-03|18:12:58.225] Started node module=main nodeInfo="{ProtocolVersion:{P2P:6 Block:9 App:1} ID_:35e87e93f2e31f305b65a5517fd2102331b56002 ListenAddr:tcp://0.0.0.0:26656 Network:test-chain-J8JvJH Version:0.29.1 Channels:4020212223303800 Moniker:Ethans-MacBook-Pro.local Other:{TxIndex:on RPCAddress:tcp://0.0.0.0:26657}}" E[2019-02-03|18:12:58.226] Couldn't connect to any seeds module=p2p I[2019-02-03|18:12:59.229] Timed out module=consensus dur=998.568ms height=1 round=0 step=RoundStepNewHeight I[2019-02-03|18:12:59.230] enterNewRound(1/0). Current: 1/0/RoundStepNewHeight module=consensus height=1 round=0 I[2019-02-03|18:12:59.230] enterPropose(1/0). Current: 1/0/RoundStepNewRound module=consensus height=1 round=0 I[2019-02-03|18:12:59.230] enterPropose: Our turn to propose module=consensus height=1 round=0 proposer=AD278B7767B05D7FBEB76207024C650988FA77D5 privValidator="PrivValidator{AD278B7767B05D7FBEB76207024C650988FA77D5 LH:1, LR:0, LS:2}" E[2019-02-03|18:12:59.230] enterPropose: Error signing proposal module=consensus height=1 round=0 err="Error signing proposal: Step regression at height 1 round 0. Got 1, last step 2" I[2019-02-03|18:13:02.233] Timed out module=consensus dur=3s height=1 round=0 step=RoundStepPropose I[2019-02-03|18:13:02.233] enterPrevote(1/0). Current: 1/0/RoundStepPropose module=consensus I[2019-02-03|18:13:02.233] enterPrevote: ProposalBlock is nil module=consensus height=1 round=0 E[2019-02-03|18:13:02.234] Error signing vote module=consensus height=1 round=0 vote="Vote{0:AD278B7767B0 1/00/1(Prevote) 000000000000 000000000000 @ 2019-02-04T02:13:02.233897Z}" err="Error signing vote: Conflicting data" ``` Notice the EOF, the step regression, and the conflicting data. * wal: change errors to be DataCorruptionError * exit on corrupt WAL * fix log * fix new line
6 years ago
WAL: better errors and new fail point (#3246) * privval: more info in errors * wal: change Debug logs to Info * wal: log and return error on corrupted wal instead of panicing * fail: Exit right away instead of sending interupt * consensus: FAIL before handling our own vote allows to replicate #3089: - run using `FAIL_TEST_INDEX=0` - delete some bytes from the end of the WAL - start normally Results in logs like: ``` I[2019-02-03|18:12:58.225] Searching for height module=consensus wal=/Users/ethanbuchman/.tendermint/data/cs.wal/wal height=1 min=0 max=0 E[2019-02-03|18:12:58.225] Error on catchup replay. Proceeding to start ConsensusState anyway module=consensus err="failed to read data: EOF" I[2019-02-03|18:12:58.225] Started node module=main nodeInfo="{ProtocolVersion:{P2P:6 Block:9 App:1} ID_:35e87e93f2e31f305b65a5517fd2102331b56002 ListenAddr:tcp://0.0.0.0:26656 Network:test-chain-J8JvJH Version:0.29.1 Channels:4020212223303800 Moniker:Ethans-MacBook-Pro.local Other:{TxIndex:on RPCAddress:tcp://0.0.0.0:26657}}" E[2019-02-03|18:12:58.226] Couldn't connect to any seeds module=p2p I[2019-02-03|18:12:59.229] Timed out module=consensus dur=998.568ms height=1 round=0 step=RoundStepNewHeight I[2019-02-03|18:12:59.230] enterNewRound(1/0). Current: 1/0/RoundStepNewHeight module=consensus height=1 round=0 I[2019-02-03|18:12:59.230] enterPropose(1/0). Current: 1/0/RoundStepNewRound module=consensus height=1 round=0 I[2019-02-03|18:12:59.230] enterPropose: Our turn to propose module=consensus height=1 round=0 proposer=AD278B7767B05D7FBEB76207024C650988FA77D5 privValidator="PrivValidator{AD278B7767B05D7FBEB76207024C650988FA77D5 LH:1, LR:0, LS:2}" E[2019-02-03|18:12:59.230] enterPropose: Error signing proposal module=consensus height=1 round=0 err="Error signing proposal: Step regression at height 1 round 0. Got 1, last step 2" I[2019-02-03|18:13:02.233] Timed out module=consensus dur=3s height=1 round=0 step=RoundStepPropose I[2019-02-03|18:13:02.233] enterPrevote(1/0). Current: 1/0/RoundStepPropose module=consensus I[2019-02-03|18:13:02.233] enterPrevote: ProposalBlock is nil module=consensus height=1 round=0 E[2019-02-03|18:13:02.234] Error signing vote module=consensus height=1 round=0 vote="Vote{0:AD278B7767B0 1/00/1(Prevote) 000000000000 000000000000 @ 2019-02-04T02:13:02.233897Z}" err="Error signing vote: Conflicting data" ``` Notice the EOF, the step regression, and the conflicting data. * wal: change errors to be DataCorruptionError * exit on corrupt WAL * fix log * fix new line
6 years ago
WAL: better errors and new fail point (#3246) * privval: more info in errors * wal: change Debug logs to Info * wal: log and return error on corrupted wal instead of panicing * fail: Exit right away instead of sending interupt * consensus: FAIL before handling our own vote allows to replicate #3089: - run using `FAIL_TEST_INDEX=0` - delete some bytes from the end of the WAL - start normally Results in logs like: ``` I[2019-02-03|18:12:58.225] Searching for height module=consensus wal=/Users/ethanbuchman/.tendermint/data/cs.wal/wal height=1 min=0 max=0 E[2019-02-03|18:12:58.225] Error on catchup replay. Proceeding to start ConsensusState anyway module=consensus err="failed to read data: EOF" I[2019-02-03|18:12:58.225] Started node module=main nodeInfo="{ProtocolVersion:{P2P:6 Block:9 App:1} ID_:35e87e93f2e31f305b65a5517fd2102331b56002 ListenAddr:tcp://0.0.0.0:26656 Network:test-chain-J8JvJH Version:0.29.1 Channels:4020212223303800 Moniker:Ethans-MacBook-Pro.local Other:{TxIndex:on RPCAddress:tcp://0.0.0.0:26657}}" E[2019-02-03|18:12:58.226] Couldn't connect to any seeds module=p2p I[2019-02-03|18:12:59.229] Timed out module=consensus dur=998.568ms height=1 round=0 step=RoundStepNewHeight I[2019-02-03|18:12:59.230] enterNewRound(1/0). Current: 1/0/RoundStepNewHeight module=consensus height=1 round=0 I[2019-02-03|18:12:59.230] enterPropose(1/0). Current: 1/0/RoundStepNewRound module=consensus height=1 round=0 I[2019-02-03|18:12:59.230] enterPropose: Our turn to propose module=consensus height=1 round=0 proposer=AD278B7767B05D7FBEB76207024C650988FA77D5 privValidator="PrivValidator{AD278B7767B05D7FBEB76207024C650988FA77D5 LH:1, LR:0, LS:2}" E[2019-02-03|18:12:59.230] enterPropose: Error signing proposal module=consensus height=1 round=0 err="Error signing proposal: Step regression at height 1 round 0. Got 1, last step 2" I[2019-02-03|18:13:02.233] Timed out module=consensus dur=3s height=1 round=0 step=RoundStepPropose I[2019-02-03|18:13:02.233] enterPrevote(1/0). Current: 1/0/RoundStepPropose module=consensus I[2019-02-03|18:13:02.233] enterPrevote: ProposalBlock is nil module=consensus height=1 round=0 E[2019-02-03|18:13:02.234] Error signing vote module=consensus height=1 round=0 vote="Vote{0:AD278B7767B0 1/00/1(Prevote) 000000000000 000000000000 @ 2019-02-04T02:13:02.233897Z}" err="Error signing vote: Conflicting data" ``` Notice the EOF, the step regression, and the conflicting data. * wal: change errors to be DataCorruptionError * exit on corrupt WAL * fix log * fix new line
6 years ago
7 years ago
7 years ago
7 years ago
WAL: better errors and new fail point (#3246) * privval: more info in errors * wal: change Debug logs to Info * wal: log and return error on corrupted wal instead of panicing * fail: Exit right away instead of sending interupt * consensus: FAIL before handling our own vote allows to replicate #3089: - run using `FAIL_TEST_INDEX=0` - delete some bytes from the end of the WAL - start normally Results in logs like: ``` I[2019-02-03|18:12:58.225] Searching for height module=consensus wal=/Users/ethanbuchman/.tendermint/data/cs.wal/wal height=1 min=0 max=0 E[2019-02-03|18:12:58.225] Error on catchup replay. Proceeding to start ConsensusState anyway module=consensus err="failed to read data: EOF" I[2019-02-03|18:12:58.225] Started node module=main nodeInfo="{ProtocolVersion:{P2P:6 Block:9 App:1} ID_:35e87e93f2e31f305b65a5517fd2102331b56002 ListenAddr:tcp://0.0.0.0:26656 Network:test-chain-J8JvJH Version:0.29.1 Channels:4020212223303800 Moniker:Ethans-MacBook-Pro.local Other:{TxIndex:on RPCAddress:tcp://0.0.0.0:26657}}" E[2019-02-03|18:12:58.226] Couldn't connect to any seeds module=p2p I[2019-02-03|18:12:59.229] Timed out module=consensus dur=998.568ms height=1 round=0 step=RoundStepNewHeight I[2019-02-03|18:12:59.230] enterNewRound(1/0). Current: 1/0/RoundStepNewHeight module=consensus height=1 round=0 I[2019-02-03|18:12:59.230] enterPropose(1/0). Current: 1/0/RoundStepNewRound module=consensus height=1 round=0 I[2019-02-03|18:12:59.230] enterPropose: Our turn to propose module=consensus height=1 round=0 proposer=AD278B7767B05D7FBEB76207024C650988FA77D5 privValidator="PrivValidator{AD278B7767B05D7FBEB76207024C650988FA77D5 LH:1, LR:0, LS:2}" E[2019-02-03|18:12:59.230] enterPropose: Error signing proposal module=consensus height=1 round=0 err="Error signing proposal: Step regression at height 1 round 0. Got 1, last step 2" I[2019-02-03|18:13:02.233] Timed out module=consensus dur=3s height=1 round=0 step=RoundStepPropose I[2019-02-03|18:13:02.233] enterPrevote(1/0). Current: 1/0/RoundStepPropose module=consensus I[2019-02-03|18:13:02.233] enterPrevote: ProposalBlock is nil module=consensus height=1 round=0 E[2019-02-03|18:13:02.234] Error signing vote module=consensus height=1 round=0 vote="Vote{0:AD278B7767B0 1/00/1(Prevote) 000000000000 000000000000 @ 2019-02-04T02:13:02.233897Z}" err="Error signing vote: Conflicting data" ``` Notice the EOF, the step regression, and the conflicting data. * wal: change errors to be DataCorruptionError * exit on corrupt WAL * fix log * fix new line
6 years ago
7 years ago
WAL: better errors and new fail point (#3246) * privval: more info in errors * wal: change Debug logs to Info * wal: log and return error on corrupted wal instead of panicing * fail: Exit right away instead of sending interupt * consensus: FAIL before handling our own vote allows to replicate #3089: - run using `FAIL_TEST_INDEX=0` - delete some bytes from the end of the WAL - start normally Results in logs like: ``` I[2019-02-03|18:12:58.225] Searching for height module=consensus wal=/Users/ethanbuchman/.tendermint/data/cs.wal/wal height=1 min=0 max=0 E[2019-02-03|18:12:58.225] Error on catchup replay. Proceeding to start ConsensusState anyway module=consensus err="failed to read data: EOF" I[2019-02-03|18:12:58.225] Started node module=main nodeInfo="{ProtocolVersion:{P2P:6 Block:9 App:1} ID_:35e87e93f2e31f305b65a5517fd2102331b56002 ListenAddr:tcp://0.0.0.0:26656 Network:test-chain-J8JvJH Version:0.29.1 Channels:4020212223303800 Moniker:Ethans-MacBook-Pro.local Other:{TxIndex:on RPCAddress:tcp://0.0.0.0:26657}}" E[2019-02-03|18:12:58.226] Couldn't connect to any seeds module=p2p I[2019-02-03|18:12:59.229] Timed out module=consensus dur=998.568ms height=1 round=0 step=RoundStepNewHeight I[2019-02-03|18:12:59.230] enterNewRound(1/0). Current: 1/0/RoundStepNewHeight module=consensus height=1 round=0 I[2019-02-03|18:12:59.230] enterPropose(1/0). Current: 1/0/RoundStepNewRound module=consensus height=1 round=0 I[2019-02-03|18:12:59.230] enterPropose: Our turn to propose module=consensus height=1 round=0 proposer=AD278B7767B05D7FBEB76207024C650988FA77D5 privValidator="PrivValidator{AD278B7767B05D7FBEB76207024C650988FA77D5 LH:1, LR:0, LS:2}" E[2019-02-03|18:12:59.230] enterPropose: Error signing proposal module=consensus height=1 round=0 err="Error signing proposal: Step regression at height 1 round 0. Got 1, last step 2" I[2019-02-03|18:13:02.233] Timed out module=consensus dur=3s height=1 round=0 step=RoundStepPropose I[2019-02-03|18:13:02.233] enterPrevote(1/0). Current: 1/0/RoundStepPropose module=consensus I[2019-02-03|18:13:02.233] enterPrevote: ProposalBlock is nil module=consensus height=1 round=0 E[2019-02-03|18:13:02.234] Error signing vote module=consensus height=1 round=0 vote="Vote{0:AD278B7767B0 1/00/1(Prevote) 000000000000 000000000000 @ 2019-02-04T02:13:02.233897Z}" err="Error signing vote: Conflicting data" ``` Notice the EOF, the step regression, and the conflicting data. * wal: change errors to be DataCorruptionError * exit on corrupt WAL * fix log * fix new line
6 years ago
WAL: better errors and new fail point (#3246) * privval: more info in errors * wal: change Debug logs to Info * wal: log and return error on corrupted wal instead of panicing * fail: Exit right away instead of sending interupt * consensus: FAIL before handling our own vote allows to replicate #3089: - run using `FAIL_TEST_INDEX=0` - delete some bytes from the end of the WAL - start normally Results in logs like: ``` I[2019-02-03|18:12:58.225] Searching for height module=consensus wal=/Users/ethanbuchman/.tendermint/data/cs.wal/wal height=1 min=0 max=0 E[2019-02-03|18:12:58.225] Error on catchup replay. Proceeding to start ConsensusState anyway module=consensus err="failed to read data: EOF" I[2019-02-03|18:12:58.225] Started node module=main nodeInfo="{ProtocolVersion:{P2P:6 Block:9 App:1} ID_:35e87e93f2e31f305b65a5517fd2102331b56002 ListenAddr:tcp://0.0.0.0:26656 Network:test-chain-J8JvJH Version:0.29.1 Channels:4020212223303800 Moniker:Ethans-MacBook-Pro.local Other:{TxIndex:on RPCAddress:tcp://0.0.0.0:26657}}" E[2019-02-03|18:12:58.226] Couldn't connect to any seeds module=p2p I[2019-02-03|18:12:59.229] Timed out module=consensus dur=998.568ms height=1 round=0 step=RoundStepNewHeight I[2019-02-03|18:12:59.230] enterNewRound(1/0). Current: 1/0/RoundStepNewHeight module=consensus height=1 round=0 I[2019-02-03|18:12:59.230] enterPropose(1/0). Current: 1/0/RoundStepNewRound module=consensus height=1 round=0 I[2019-02-03|18:12:59.230] enterPropose: Our turn to propose module=consensus height=1 round=0 proposer=AD278B7767B05D7FBEB76207024C650988FA77D5 privValidator="PrivValidator{AD278B7767B05D7FBEB76207024C650988FA77D5 LH:1, LR:0, LS:2}" E[2019-02-03|18:12:59.230] enterPropose: Error signing proposal module=consensus height=1 round=0 err="Error signing proposal: Step regression at height 1 round 0. Got 1, last step 2" I[2019-02-03|18:13:02.233] Timed out module=consensus dur=3s height=1 round=0 step=RoundStepPropose I[2019-02-03|18:13:02.233] enterPrevote(1/0). Current: 1/0/RoundStepPropose module=consensus I[2019-02-03|18:13:02.233] enterPrevote: ProposalBlock is nil module=consensus height=1 round=0 E[2019-02-03|18:13:02.234] Error signing vote module=consensus height=1 round=0 vote="Vote{0:AD278B7767B0 1/00/1(Prevote) 000000000000 000000000000 @ 2019-02-04T02:13:02.233897Z}" err="Error signing vote: Conflicting data" ``` Notice the EOF, the step regression, and the conflicting data. * wal: change errors to be DataCorruptionError * exit on corrupt WAL * fix log * fix new line
6 years ago
7 years ago
WAL: better errors and new fail point (#3246) * privval: more info in errors * wal: change Debug logs to Info * wal: log and return error on corrupted wal instead of panicing * fail: Exit right away instead of sending interupt * consensus: FAIL before handling our own vote allows to replicate #3089: - run using `FAIL_TEST_INDEX=0` - delete some bytes from the end of the WAL - start normally Results in logs like: ``` I[2019-02-03|18:12:58.225] Searching for height module=consensus wal=/Users/ethanbuchman/.tendermint/data/cs.wal/wal height=1 min=0 max=0 E[2019-02-03|18:12:58.225] Error on catchup replay. Proceeding to start ConsensusState anyway module=consensus err="failed to read data: EOF" I[2019-02-03|18:12:58.225] Started node module=main nodeInfo="{ProtocolVersion:{P2P:6 Block:9 App:1} ID_:35e87e93f2e31f305b65a5517fd2102331b56002 ListenAddr:tcp://0.0.0.0:26656 Network:test-chain-J8JvJH Version:0.29.1 Channels:4020212223303800 Moniker:Ethans-MacBook-Pro.local Other:{TxIndex:on RPCAddress:tcp://0.0.0.0:26657}}" E[2019-02-03|18:12:58.226] Couldn't connect to any seeds module=p2p I[2019-02-03|18:12:59.229] Timed out module=consensus dur=998.568ms height=1 round=0 step=RoundStepNewHeight I[2019-02-03|18:12:59.230] enterNewRound(1/0). Current: 1/0/RoundStepNewHeight module=consensus height=1 round=0 I[2019-02-03|18:12:59.230] enterPropose(1/0). Current: 1/0/RoundStepNewRound module=consensus height=1 round=0 I[2019-02-03|18:12:59.230] enterPropose: Our turn to propose module=consensus height=1 round=0 proposer=AD278B7767B05D7FBEB76207024C650988FA77D5 privValidator="PrivValidator{AD278B7767B05D7FBEB76207024C650988FA77D5 LH:1, LR:0, LS:2}" E[2019-02-03|18:12:59.230] enterPropose: Error signing proposal module=consensus height=1 round=0 err="Error signing proposal: Step regression at height 1 round 0. Got 1, last step 2" I[2019-02-03|18:13:02.233] Timed out module=consensus dur=3s height=1 round=0 step=RoundStepPropose I[2019-02-03|18:13:02.233] enterPrevote(1/0). Current: 1/0/RoundStepPropose module=consensus I[2019-02-03|18:13:02.233] enterPrevote: ProposalBlock is nil module=consensus height=1 round=0 E[2019-02-03|18:13:02.234] Error signing vote module=consensus height=1 round=0 vote="Vote{0:AD278B7767B0 1/00/1(Prevote) 000000000000 000000000000 @ 2019-02-04T02:13:02.233897Z}" err="Error signing vote: Conflicting data" ``` Notice the EOF, the step regression, and the conflicting data. * wal: change errors to be DataCorruptionError * exit on corrupt WAL * fix log * fix new line
6 years ago
7 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
7 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
7 years ago
  1. package consensus
  2. import (
  3. "encoding/binary"
  4. "fmt"
  5. "hash/crc32"
  6. "io"
  7. "path/filepath"
  8. "time"
  9. "github.com/pkg/errors"
  10. amino "github.com/tendermint/go-amino"
  11. auto "github.com/tendermint/tendermint/libs/autofile"
  12. cmn "github.com/tendermint/tendermint/libs/common"
  13. "github.com/tendermint/tendermint/libs/log"
  14. "github.com/tendermint/tendermint/types"
  15. tmtime "github.com/tendermint/tendermint/types/time"
  16. )
  17. const (
  18. // must be greater than types.BlockPartSizeBytes + a few bytes
  19. maxMsgSizeBytes = 1024 * 1024 // 1MB
  20. )
  21. //--------------------------------------------------------
  22. // types and functions for savings consensus messages
  23. type TimedWALMessage struct {
  24. Time time.Time `json:"time"` // for debugging purposes
  25. Msg WALMessage `json:"msg"`
  26. }
  27. // EndHeightMessage marks the end of the given height inside WAL.
  28. // @internal used by scripts/wal2json util.
  29. type EndHeightMessage struct {
  30. Height int64 `json:"height"`
  31. }
  32. type WALMessage interface{}
  33. func RegisterWALMessages(cdc *amino.Codec) {
  34. cdc.RegisterInterface((*WALMessage)(nil), nil)
  35. cdc.RegisterConcrete(types.EventDataRoundState{}, "tendermint/wal/EventDataRoundState", nil)
  36. cdc.RegisterConcrete(msgInfo{}, "tendermint/wal/MsgInfo", nil)
  37. cdc.RegisterConcrete(timeoutInfo{}, "tendermint/wal/TimeoutInfo", nil)
  38. cdc.RegisterConcrete(EndHeightMessage{}, "tendermint/wal/EndHeightMessage", nil)
  39. }
  40. //--------------------------------------------------------
  41. // Simple write-ahead logger
  42. // WAL is an interface for any write-ahead logger.
  43. type WAL interface {
  44. Write(WALMessage)
  45. WriteSync(WALMessage)
  46. Group() *auto.Group
  47. SearchForEndHeight(height int64, options *WALSearchOptions) (gr *auto.GroupReader, found bool, err error)
  48. Start() error
  49. Stop() error
  50. Wait()
  51. }
  52. // Write ahead logger writes msgs to disk before they are processed.
  53. // Can be used for crash-recovery and deterministic replay
  54. // TODO: currently the wal is overwritten during replay catchup
  55. // give it a mode so it's either reading or appending - must read to end to start appending again
  56. type baseWAL struct {
  57. cmn.BaseService
  58. group *auto.Group
  59. enc *WALEncoder
  60. }
  61. func NewWAL(walFile string, groupOptions ...func(*auto.Group)) (*baseWAL, error) {
  62. err := cmn.EnsureDir(filepath.Dir(walFile), 0700)
  63. if err != nil {
  64. return nil, errors.Wrap(err, "failed to ensure WAL directory is in place")
  65. }
  66. group, err := auto.OpenGroup(walFile, groupOptions...)
  67. if err != nil {
  68. return nil, err
  69. }
  70. wal := &baseWAL{
  71. group: group,
  72. enc: NewWALEncoder(group),
  73. }
  74. wal.BaseService = *cmn.NewBaseService(nil, "baseWAL", wal)
  75. return wal, nil
  76. }
  77. func (wal *baseWAL) Group() *auto.Group {
  78. return wal.group
  79. }
  80. func (wal *baseWAL) SetLogger(l log.Logger) {
  81. wal.BaseService.Logger = l
  82. wal.group.SetLogger(l)
  83. }
  84. func (wal *baseWAL) OnStart() error {
  85. size, err := wal.group.Head.Size()
  86. if err != nil {
  87. return err
  88. } else if size == 0 {
  89. wal.WriteSync(EndHeightMessage{0})
  90. }
  91. err = wal.group.Start()
  92. return err
  93. }
  94. // Stop the underlying autofile group.
  95. // Use Wait() to ensure it's finished shutting down
  96. // before cleaning up files.
  97. func (wal *baseWAL) OnStop() {
  98. wal.group.Stop()
  99. wal.group.Close()
  100. }
  101. // Wait for the underlying autofile group to finish shutting down
  102. // so it's safe to cleanup files.
  103. func (wal *baseWAL) Wait() {
  104. wal.group.Wait()
  105. }
  106. // Write is called in newStep and for each receive on the
  107. // peerMsgQueue and the timeoutTicker.
  108. // NOTE: does not call fsync()
  109. func (wal *baseWAL) Write(msg WALMessage) {
  110. if wal == nil {
  111. return
  112. }
  113. // Write the wal message
  114. if err := wal.enc.Encode(&TimedWALMessage{tmtime.Now(), msg}); err != nil {
  115. panic(fmt.Sprintf("Error writing msg to consensus wal: %v \n\nMessage: %v", err, msg))
  116. }
  117. }
  118. // WriteSync is called when we receive a msg from ourselves
  119. // so that we write to disk before sending signed messages.
  120. // NOTE: calls fsync()
  121. func (wal *baseWAL) WriteSync(msg WALMessage) {
  122. if wal == nil {
  123. return
  124. }
  125. wal.Write(msg)
  126. if err := wal.group.Flush(); err != nil {
  127. panic(fmt.Sprintf("Error flushing consensus wal buf to file. Error: %v \n", err))
  128. }
  129. }
  130. // WALSearchOptions are optional arguments to SearchForEndHeight.
  131. type WALSearchOptions struct {
  132. // IgnoreDataCorruptionErrors set to true will result in skipping data corruption errors.
  133. IgnoreDataCorruptionErrors bool
  134. }
  135. // SearchForEndHeight searches for the EndHeightMessage with the given height
  136. // and returns an auto.GroupReader, whenever it was found or not and an error.
  137. // Group reader will be nil if found equals false.
  138. //
  139. // CONTRACT: caller must close group reader.
  140. func (wal *baseWAL) SearchForEndHeight(height int64, options *WALSearchOptions) (gr *auto.GroupReader, found bool, err error) {
  141. var msg *TimedWALMessage
  142. lastHeightFound := int64(-1)
  143. // NOTE: starting from the last file in the group because we're usually
  144. // searching for the last height. See replay.go
  145. min, max := wal.group.MinIndex(), wal.group.MaxIndex()
  146. wal.Logger.Info("Searching for height", "height", height, "min", min, "max", max)
  147. for index := max; index >= min; index-- {
  148. gr, err = wal.group.NewReader(index)
  149. if err != nil {
  150. return nil, false, err
  151. }
  152. dec := NewWALDecoder(gr)
  153. for {
  154. msg, err = dec.Decode()
  155. if err == io.EOF {
  156. // OPTIMISATION: no need to look for height in older files if we've seen h < height
  157. if lastHeightFound > 0 && lastHeightFound < height {
  158. gr.Close()
  159. return nil, false, nil
  160. }
  161. // check next file
  162. break
  163. }
  164. if options.IgnoreDataCorruptionErrors && IsDataCorruptionError(err) {
  165. wal.Logger.Error("Corrupted entry. Skipping...", "err", err)
  166. // do nothing
  167. continue
  168. } else if err != nil {
  169. gr.Close()
  170. return nil, false, err
  171. }
  172. if m, ok := msg.Msg.(EndHeightMessage); ok {
  173. lastHeightFound = m.Height
  174. if m.Height == height { // found
  175. wal.Logger.Info("Found", "height", height, "index", index)
  176. return gr, true, nil
  177. }
  178. }
  179. }
  180. gr.Close()
  181. }
  182. return nil, false, nil
  183. }
  184. ///////////////////////////////////////////////////////////////////////////////
  185. // A WALEncoder writes custom-encoded WAL messages to an output stream.
  186. //
  187. // Format: 4 bytes CRC sum + 4 bytes length + arbitrary-length value (go-amino encoded)
  188. type WALEncoder struct {
  189. wr io.Writer
  190. }
  191. // NewWALEncoder returns a new encoder that writes to wr.
  192. func NewWALEncoder(wr io.Writer) *WALEncoder {
  193. return &WALEncoder{wr}
  194. }
  195. // Encode writes the custom encoding of v to the stream.
  196. func (enc *WALEncoder) Encode(v *TimedWALMessage) error {
  197. data := cdc.MustMarshalBinaryBare(v)
  198. crc := crc32.Checksum(data, crc32c)
  199. length := uint32(len(data))
  200. totalLength := 8 + int(length)
  201. msg := make([]byte, totalLength)
  202. binary.BigEndian.PutUint32(msg[0:4], crc)
  203. binary.BigEndian.PutUint32(msg[4:8], length)
  204. copy(msg[8:], data)
  205. _, err := enc.wr.Write(msg)
  206. return err
  207. }
  208. ///////////////////////////////////////////////////////////////////////////////
  209. // IsDataCorruptionError returns true if data has been corrupted inside WAL.
  210. func IsDataCorruptionError(err error) bool {
  211. _, ok := err.(DataCorruptionError)
  212. return ok
  213. }
  214. // DataCorruptionError is an error that occures if data on disk was corrupted.
  215. type DataCorruptionError struct {
  216. cause error
  217. }
  218. func (e DataCorruptionError) Error() string {
  219. return fmt.Sprintf("DataCorruptionError[%v]", e.cause)
  220. }
  221. func (e DataCorruptionError) Cause() error {
  222. return e.cause
  223. }
  224. // A WALDecoder reads and decodes custom-encoded WAL messages from an input
  225. // stream. See WALEncoder for the format used.
  226. //
  227. // It will also compare the checksums and make sure data size is equal to the
  228. // length from the header. If that is not the case, error will be returned.
  229. type WALDecoder struct {
  230. rd io.Reader
  231. }
  232. // NewWALDecoder returns a new decoder that reads from rd.
  233. func NewWALDecoder(rd io.Reader) *WALDecoder {
  234. return &WALDecoder{rd}
  235. }
  236. // Decode reads the next custom-encoded value from its reader and returns it.
  237. func (dec *WALDecoder) Decode() (*TimedWALMessage, error) {
  238. b := make([]byte, 4)
  239. _, err := dec.rd.Read(b)
  240. if err == io.EOF {
  241. return nil, err
  242. }
  243. if err != nil {
  244. return nil, DataCorruptionError{fmt.Errorf("failed to read checksum: %v", err)}
  245. }
  246. crc := binary.BigEndian.Uint32(b)
  247. b = make([]byte, 4)
  248. _, err = dec.rd.Read(b)
  249. if err != nil {
  250. return nil, DataCorruptionError{fmt.Errorf("failed to read length: %v", err)}
  251. }
  252. length := binary.BigEndian.Uint32(b)
  253. if length > maxMsgSizeBytes {
  254. return nil, DataCorruptionError{fmt.Errorf("length %d exceeded maximum possible value of %d bytes", length, maxMsgSizeBytes)}
  255. }
  256. data := make([]byte, length)
  257. _, err = dec.rd.Read(data)
  258. if err != nil {
  259. return nil, DataCorruptionError{fmt.Errorf("failed to read data: %v", err)}
  260. }
  261. // check checksum before decoding data
  262. actualCRC := crc32.Checksum(data, crc32c)
  263. if actualCRC != crc {
  264. return nil, DataCorruptionError{fmt.Errorf("checksums do not match: (read: %v, actual: %v)", crc, actualCRC)}
  265. }
  266. var res = new(TimedWALMessage) // nolint: gosimple
  267. err = cdc.UnmarshalBinaryBare(data, res)
  268. if err != nil {
  269. return nil, DataCorruptionError{fmt.Errorf("failed to decode data: %v", err)}
  270. }
  271. return res, err
  272. }
  273. type nilWAL struct{}
  274. func (nilWAL) Write(m WALMessage) {}
  275. func (nilWAL) WriteSync(m WALMessage) {}
  276. func (nilWAL) Group() *auto.Group { return nil }
  277. func (nilWAL) SearchForEndHeight(height int64, options *WALSearchOptions) (gr *auto.GroupReader, found bool, err error) {
  278. return nil, false, nil
  279. }
  280. func (nilWAL) Start() error { return nil }
  281. func (nilWAL) Stop() error { return nil }
  282. func (nilWAL) Wait() {}