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.

324 lines
8.8 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
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
7 years ago
7 years ago
7 years ago
7 years ago
8 years ago
7 years ago
7 years ago
7 years ago
7 years ago
7 years ago
7 years ago
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
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. "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/types"
  14. tmtime "github.com/tendermint/tendermint/types/time"
  15. )
  16. const (
  17. // must be greater than params.BlockGossip.BlockPartSizeBytes + a few bytes
  18. maxMsgSizeBytes = 1024 * 1024 // 1MB
  19. )
  20. //--------------------------------------------------------
  21. // types and functions for savings consensus messages
  22. type TimedWALMessage struct {
  23. Time time.Time `json:"time"` // for debugging purposes
  24. Msg WALMessage `json:"msg"`
  25. }
  26. // EndHeightMessage marks the end of the given height inside WAL.
  27. // @internal used by scripts/wal2json util.
  28. type EndHeightMessage struct {
  29. Height int64 `json:"height"`
  30. }
  31. type WALMessage interface{}
  32. func RegisterWALMessages(cdc *amino.Codec) {
  33. cdc.RegisterInterface((*WALMessage)(nil), nil)
  34. cdc.RegisterConcrete(types.EventDataRoundState{}, "tendermint/wal/EventDataRoundState", nil)
  35. cdc.RegisterConcrete(msgInfo{}, "tendermint/wal/MsgInfo", nil)
  36. cdc.RegisterConcrete(timeoutInfo{}, "tendermint/wal/TimeoutInfo", nil)
  37. cdc.RegisterConcrete(EndHeightMessage{}, "tendermint/wal/EndHeightMessage", nil)
  38. }
  39. //--------------------------------------------------------
  40. // Simple write-ahead logger
  41. // WAL is an interface for any write-ahead logger.
  42. type WAL interface {
  43. Write(WALMessage)
  44. WriteSync(WALMessage)
  45. Group() *auto.Group
  46. SearchForEndHeight(height int64, options *WALSearchOptions) (gr *auto.GroupReader, found bool, err error)
  47. Start() error
  48. Stop() error
  49. Wait()
  50. }
  51. // Write ahead logger writes msgs to disk before they are processed.
  52. // Can be used for crash-recovery and deterministic replay
  53. // TODO: currently the wal is overwritten during replay catchup
  54. // give it a mode so it's either reading or appending - must read to end to start appending again
  55. type baseWAL struct {
  56. cmn.BaseService
  57. group *auto.Group
  58. enc *WALEncoder
  59. }
  60. func NewWAL(walFile string) (*baseWAL, error) {
  61. err := cmn.EnsureDir(filepath.Dir(walFile), 0700)
  62. if err != nil {
  63. return nil, errors.Wrap(err, "failed to ensure WAL directory is in place")
  64. }
  65. group, err := auto.OpenGroup(walFile)
  66. if err != nil {
  67. return nil, err
  68. }
  69. wal := &baseWAL{
  70. group: group,
  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.WriteSync(EndHeightMessage{0})
  85. }
  86. err = wal.group.Start()
  87. return err
  88. }
  89. func (wal *baseWAL) OnStop() {
  90. wal.group.Stop()
  91. wal.group.Close()
  92. }
  93. // Write is called in newStep and for each receive on the
  94. // peerMsgQueue and the timeoutTicker.
  95. // NOTE: does not call fsync()
  96. func (wal *baseWAL) Write(msg WALMessage) {
  97. if wal == nil {
  98. return
  99. }
  100. // Write the wal message
  101. if err := wal.enc.Encode(&TimedWALMessage{tmtime.Now(), msg}); err != nil {
  102. panic(fmt.Sprintf("Error writing msg to consensus wal: %v \n\nMessage: %v", err, msg))
  103. }
  104. }
  105. // WriteSync is called when we receive a msg from ourselves
  106. // so that we write to disk before sending signed messages.
  107. // NOTE: calls fsync()
  108. func (wal *baseWAL) WriteSync(msg WALMessage) {
  109. if wal == nil {
  110. return
  111. }
  112. wal.Write(msg)
  113. if err := wal.group.Flush(); err != nil {
  114. panic(fmt.Sprintf("Error flushing consensus wal buf to file. Error: %v \n", err))
  115. }
  116. }
  117. // WALSearchOptions are optional arguments to SearchForEndHeight.
  118. type WALSearchOptions struct {
  119. // IgnoreDataCorruptionErrors set to true will result in skipping data corruption errors.
  120. IgnoreDataCorruptionErrors bool
  121. }
  122. // SearchForEndHeight searches for the EndHeightMessage with the given height
  123. // and returns an auto.GroupReader, whenever it was found or not and an error.
  124. // Group reader will be nil if found equals false.
  125. //
  126. // CONTRACT: caller must close group reader.
  127. func (wal *baseWAL) SearchForEndHeight(height int64, options *WALSearchOptions) (gr *auto.GroupReader, found bool, err error) {
  128. var msg *TimedWALMessage
  129. lastHeightFound := int64(-1)
  130. // NOTE: starting from the last file in the group because we're usually
  131. // searching for the last height. See replay.go
  132. min, max := wal.group.MinIndex(), wal.group.MaxIndex()
  133. wal.Logger.Debug("Searching for height", "height", height, "min", min, "max", max)
  134. for index := max; index >= min; index-- {
  135. gr, err = wal.group.NewReader(index)
  136. if err != nil {
  137. return nil, false, err
  138. }
  139. dec := NewWALDecoder(gr)
  140. for {
  141. msg, err = dec.Decode()
  142. if err == io.EOF {
  143. // OPTIMISATION: no need to look for height in older files if we've seen h < height
  144. if lastHeightFound > 0 && lastHeightFound < height {
  145. gr.Close()
  146. return nil, false, nil
  147. }
  148. // check next file
  149. break
  150. }
  151. if options.IgnoreDataCorruptionErrors && IsDataCorruptionError(err) {
  152. wal.Logger.Debug("Corrupted entry. Skipping...", "err", err)
  153. // do nothing
  154. continue
  155. } else if err != nil {
  156. gr.Close()
  157. return nil, false, err
  158. }
  159. if m, ok := msg.Msg.(EndHeightMessage); ok {
  160. lastHeightFound = m.Height
  161. if m.Height == height { // found
  162. wal.Logger.Debug("Found", "height", height, "index", index)
  163. return gr, true, nil
  164. }
  165. }
  166. }
  167. gr.Close()
  168. }
  169. return nil, false, nil
  170. }
  171. ///////////////////////////////////////////////////////////////////////////////
  172. // A WALEncoder writes custom-encoded WAL messages to an output stream.
  173. //
  174. // Format: 4 bytes CRC sum + 4 bytes length + arbitrary-length value (go-amino encoded)
  175. type WALEncoder struct {
  176. wr io.Writer
  177. }
  178. // NewWALEncoder returns a new encoder that writes to wr.
  179. func NewWALEncoder(wr io.Writer) *WALEncoder {
  180. return &WALEncoder{wr}
  181. }
  182. // Encode writes the custom encoding of v to the stream.
  183. func (enc *WALEncoder) Encode(v *TimedWALMessage) error {
  184. data := cdc.MustMarshalBinaryBare(v)
  185. crc := crc32.Checksum(data, crc32c)
  186. length := uint32(len(data))
  187. totalLength := 8 + int(length)
  188. msg := make([]byte, totalLength)
  189. binary.BigEndian.PutUint32(msg[0:4], crc)
  190. binary.BigEndian.PutUint32(msg[4:8], length)
  191. copy(msg[8:], data)
  192. _, err := enc.wr.Write(msg)
  193. return err
  194. }
  195. ///////////////////////////////////////////////////////////////////////////////
  196. // IsDataCorruptionError returns true if data has been corrupted inside WAL.
  197. func IsDataCorruptionError(err error) bool {
  198. _, ok := err.(DataCorruptionError)
  199. return ok
  200. }
  201. // DataCorruptionError is an error that occures if data on disk was corrupted.
  202. type DataCorruptionError struct {
  203. cause error
  204. }
  205. func (e DataCorruptionError) Error() string {
  206. return fmt.Sprintf("DataCorruptionError[%v]", e.cause)
  207. }
  208. func (e DataCorruptionError) Cause() error {
  209. return e.cause
  210. }
  211. // A WALDecoder reads and decodes custom-encoded WAL messages from an input
  212. // stream. See WALEncoder for the format used.
  213. //
  214. // It will also compare the checksums and make sure data size is equal to the
  215. // length from the header. If that is not the case, error will be returned.
  216. type WALDecoder struct {
  217. rd io.Reader
  218. }
  219. // NewWALDecoder returns a new decoder that reads from rd.
  220. func NewWALDecoder(rd io.Reader) *WALDecoder {
  221. return &WALDecoder{rd}
  222. }
  223. // Decode reads the next custom-encoded value from its reader and returns it.
  224. func (dec *WALDecoder) Decode() (*TimedWALMessage, error) {
  225. b := make([]byte, 4)
  226. _, err := dec.rd.Read(b)
  227. if err == io.EOF {
  228. return nil, err
  229. }
  230. if err != nil {
  231. return nil, fmt.Errorf("failed to read checksum: %v", err)
  232. }
  233. crc := binary.BigEndian.Uint32(b)
  234. b = make([]byte, 4)
  235. _, err = dec.rd.Read(b)
  236. if err != nil {
  237. return nil, fmt.Errorf("failed to read length: %v", err)
  238. }
  239. length := binary.BigEndian.Uint32(b)
  240. if length > maxMsgSizeBytes {
  241. return nil, fmt.Errorf("length %d exceeded maximum possible value of %d bytes", length, maxMsgSizeBytes)
  242. }
  243. data := make([]byte, length)
  244. _, err = dec.rd.Read(data)
  245. if err != nil {
  246. return nil, fmt.Errorf("failed to read data: %v", err)
  247. }
  248. // check checksum before decoding data
  249. actualCRC := crc32.Checksum(data, crc32c)
  250. if actualCRC != crc {
  251. return nil, DataCorruptionError{fmt.Errorf("checksums do not match: (read: %v, actual: %v)", crc, actualCRC)}
  252. }
  253. var res = new(TimedWALMessage) // nolint: gosimple
  254. err = cdc.UnmarshalBinaryBare(data, res)
  255. if err != nil {
  256. return nil, DataCorruptionError{fmt.Errorf("failed to decode data: %v", err)}
  257. }
  258. return res, err
  259. }
  260. type nilWAL struct{}
  261. func (nilWAL) Write(m WALMessage) {}
  262. func (nilWAL) WriteSync(m WALMessage) {}
  263. func (nilWAL) Group() *auto.Group { return nil }
  264. func (nilWAL) SearchForEndHeight(height int64, options *WALSearchOptions) (gr *auto.GroupReader, found bool, err error) {
  265. return nil, false, nil
  266. }
  267. func (nilWAL) Start() error { return nil }
  268. func (nilWAL) Stop() error { return nil }
  269. func (nilWAL) Wait() {}