fd35c48 | JingsongLi | 15 July 2021, 02:07:10 UTC | Commit for release 1.12.5 | 15 July 2021, 02:07:10 UTC |
8f0e13c | Stephan Ewen | 13 July 2021, 15:01:39 UTC | [FLINK-22547][tests] Harden OperatorCoordinatorHolderTest. Ensure that the 'FutureCompletedAfterSendingEventsCoordinator' cannot exit before it has completed the triggered checkpoint (completed the checkpoint future). | 14 July 2021, 14:16:01 UTC |
adfd85f | Nico Kruber | 16 June 2021, 16:16:32 UTC | [FLINK-23312][ci] speed up compilation for e2e tests The "compile" builder already applies all checks so we can use -Dfast here; also, the web UI is not actually needed in the E2E tests. | 14 July 2021, 08:07:57 UTC |
331ae2c | Stephan Ewen | 13 July 2021, 12:36:32 UTC | [FLINK-22545][coordination] Fix check during creation of Source Coordinator thread. The check was meant as a safeguard to prevent re-instantiation after fatal errors killed a previous thread. But the check was susceptible to thread termination due to idleness in the executor. This updates the check to only fail if there is in fact an instantiation next to a running thread, or after a previously crashed thread. | 13 July 2021, 22:31:43 UTC |
2c3f8f6 | Yun Gao | 09 July 2021, 23:06:22 UTC | [FLINK-23233][runtime] Ensure checkpoints confirmed after all the failed events processed for OepratorCoordinator This closes #16432. | 13 July 2021, 12:30:16 UTC |
6292777 | Yun Gao | 04 July 2021, 14:17:42 UTC | [FLINK-23223] Notifies if there are available data on resumption for pipelined subpartition | 12 July 2021, 09:55:20 UTC |
86cc6a1 | tsreaper | 12 July 2021, 08:25:09 UTC | [FLINK-23184][table-runtime-blink] Fix compile error in code generation of unary plus and minus This closes #16435 | 12 July 2021, 08:25:09 UTC |
da53f93 | Till Rohrmann | 09 July 2021, 10:14:56 UTC | [FLINK-23076][tests] Harden DispatcherTest.testWaitingForJobMasterLeadership The test DispatcherTest.testWaitingForJobMasterLeadership is unstable because it uses a 10 ms timeout to decide whether the JobMasterService has been created or not. On our CI infrastructure this 10 ms can be too short to complete the future with JobStatus.INITIALIZING which indicates that the JobMasterService has not been created. This commit hardens the test by increasing the timeout to 200 ms and to increase the number of retries from 5 to 50. This closes #16444. | 12 July 2021, 07:56:27 UTC |
e599aa2 | Till Rohrmann | 09 July 2021, 09:33:47 UTC | [FLINK-23164][tests] Harden JobMasterTest.testMultipleStartsWork The JobMasterTest.testMultipleStartsWork fails because we don't wait for the JobMaster to finish its suspension before restarting the JobMaster. This can lead to ignoring the underlying control START control message. This again can lead to a situation where we are in the STOPPED state and try to send UnfencedMessage(startJobExecution) in JobMaster.start. This commit fixes this problem by waiting on the suspension future to complete before restarting the JobMaster. This closes #16442. | 12 July 2021, 07:53:25 UTC |
2eaabca | hehuiyuan | 12 July 2021, 03:47:41 UTC | [FLINK-20321][formats] Fix NPE when using Avro/Json/Csv formats to deserialize null input (#16394) Co-authored-by: Xue Wang <xzwxpop@hotmail.com> | 12 July 2021, 03:47:41 UTC |
e280421 | David Moravek | 07 July 2021, 13:27:07 UTC | [FLINK-22819][yarn] Remove 'yarn.am.liveness-monitor.expiry-interval-ms' override for tests. This is an interval between AM container allocation and actually running RM Client inside this container, which can take longer in resource limited environment such as CI. If heartbeats between AM and RM don't work and a test relies on this, then it will fail a bit later because the default value is 5 minutes. This closes #16413. | 09 July 2021, 14:04:20 UTC |
09ac6e5 | Ada Wong | 02 July 2021, 08:06:09 UTC | [FLINK-23074][connector-hive] Shade parquet class in hive-exec to prevent conflict with flink-parquet module This closes #16424 | 09 July 2021, 04:17:17 UTC |
4812b42 | TsReaper | 05 July 2021, 07:31:32 UTC | [FLINK-22443][streaming-java] Fix overflow in MultipleInputSelectionHandler | 08 July 2021, 06:33:49 UTC |
bb29f5e | Matthias Pohl | 06 July 2021, 07:29:47 UTC | [FLINK-21445][clients] Adds configuration to ClassPathPackagedProgramRetriever | 07 July 2021, 18:11:08 UTC |
4ef1427 | Roman Khachatryan | 07 April 2021, 09:48:02 UTC | [hotfix][tests] Close SubtaskCheckpointCoordinator in tests This is to reveal any async failures occurring for example when writing channel state. | 06 July 2021, 20:45:16 UTC |
ae24c97 | Roman Khachatryan | 07 April 2021, 11:03:41 UTC | [FLINK-22105][tests] Fix checkpoint id in testForceAlignedCheckpointResultingInPriorityEvents | 06 July 2021, 20:45:16 UTC |
4d017d3 | Michal Ciesielczyk | 29 June 2021, 22:23:42 UTC | [FLINK-23182][connectors/rabbitmq] Fix connection leak in RMQSource | 06 July 2021, 16:45:36 UTC |
701d20e | Fabian Paul | 05 July 2021, 08:59:48 UTC | [FLINK-23248][datastream] Close SinkWriter when calling dispose Without this change the SinkWriter was only closed when the operator finishes and not during failing. Now, the SinkWriter is closed on dispose which is called on finish and on failure. | 06 July 2021, 08:15:20 UTC |
5186060 | Robert Metzger | 25 June 2021, 07:22:50 UTC | [FLINK-23052][ci] Improve stability of maven snapshot deployment This closes #16331 | 30 June 2021, 09:30:55 UTC |
83a26d5 | Arvid Heise | 23 June 2021, 15:08:16 UTC | [FLINK-20888][runtime] Close outputs in OperatorChain. OperatorChain creates the outputs and owns them, so that it should also close them. Specific operators should not close the outputs. Also, ChainingOutput should never close the chained operator; it's not owning the operator. | 30 June 2021, 07:49:19 UTC |
2a8cc56 | Arvid Heise | 23 June 2021, 15:16:17 UTC | [hotfix][datastream] Remove raw casts in ContinuousFileReaderOperator. | 30 June 2021, 07:49:19 UTC |
ca999b2 | Arvid Heise | 29 June 2021, 13:00:32 UTC | [FLINK-22964][connector/common] Exclude flink-core from connector-common dependencies. Connectors get shaded into the user jar and as such should contain no unnecessary dependencies to flink. However, connector-base is exposing `flink-core` which then by default gets shaded into the user jar. Except for 6MB of extra size, the dependency also causes class loading issues, when `classloader.parent-first-patterns` does not include `o.a.f`. | 30 June 2021, 07:44:01 UTC |
ab04b11 | Danny Cranmer | 18 June 2021, 13:33:55 UTC | [FLINK-18182][kinesis] Updating to latest AWS SDK for Kinesis connector | 29 June 2021, 07:25:55 UTC |
d1b8c5f | TsReaper | 09 April 2021, 07:39:34 UTC | [FLINK-22157][table-planner-blink] Fix join & select a portion of composite primary key will cause ArrayIndexOutOfBoundsException This closes #16305 (cherry picked from commit 92fbe7f1fe5f0eade036b4184cdbab8f9b791647) | 29 June 2021, 02:37:29 UTC |
169c967 | Dian Fu | 28 June 2021, 12:14:51 UTC | [FLINK-23166][python] Fix ZipUtils to handle properly for softlinks This closes #16309. | 28 June 2021, 23:15:59 UTC |
b577808 | Emre Kartoglu | 27 June 2021, 09:56:47 UTC | [FLINK-23009][kinesis] Include guava:failureaccess This wasn't included in the flink-sql-connector-kinesis jar and this caused an e2e test failure in another release branch. The other release branch is fixed in PR 16297. Applying the same changes here. | 28 June 2021, 10:20:19 UTC |
91d9004 | Dian Fu | 24 June 2021, 07:26:11 UTC | [FLINK-23133][python] Properly handle the dependencies when mixing use of Python Table API and Python DataStream API This closes #16272. | 27 June 2021, 07:07:28 UTC |
d77c51d | Dian Fu | 23 June 2021, 09:20:20 UTC | [FLINK-23120][python] Fix ByteArrayWrapperSerializer.serialize to use writeInt to serialize the length This closes #16258. | 27 June 2021, 02:14:29 UTC |
f3c22a8 | Till Rohrmann | 25 June 2021, 15:39:21 UTC | [hotfix][tests] Correct variable name in RunnablesTest.testExecutorService_uncaughtExceptionHandler | 25 June 2021, 15:48:02 UTC |
872a45b | Till Rohrmann | 23 June 2021, 14:26:23 UTC | [FLINK-23045][tests] Harden RunnablesTest by not relying on timeout This commit hardens the RunnablesTest.testExecutorService_uncaughtExceptionHandler to not rely on timeouts to check whether the uncaught exception handler was called. This closes #16263. | 25 June 2021, 15:47:45 UTC |
8861251 | Till Rohrmann | 23 June 2021, 14:24:23 UTC | [hotfix] Let RunnablesTest extend TestLogger | 25 June 2021, 15:47:39 UTC |
3a7552a | Rui Li | 17 June 2021, 08:49:51 UTC | [FLINK-23010][hive] HivePartitionFetcherContextBase shouldn't list folders to discover new partitions This closes #16182 | 25 June 2021, 06:55:27 UTC |
1619bbb | huangxingbo | 24 June 2021, 03:55:46 UTC | [FLINK-23119][python] Throw exceptions when compiling the job at places where Python UDAF is not supported This is to give a more helpful message if Python UDAF is used in places where it's not supported. This closes #16270. | 24 June 2021, 12:13:35 UTC |
a21baa4 | Timo Walther | 23 June 2021, 06:40:56 UTC | [FLINK-21229][docs] Update avro-confluent docs This closes #15808. | 24 June 2021, 06:57:14 UTC |
5bafade | Lars Bachmann | 29 April 2021, 07:31:51 UTC | [FLINK-21229][avro-confluent-registry] Add Confluent schema registry SSL support | 24 June 2021, 06:57:14 UTC |
cdec22e | huangxingbo | 11 June 2021, 07:55:17 UTC | [FLINK-22927][python] Fix the bug of JobStatus This closes #16146. | 24 June 2021, 03:01:14 UTC |
20da2a1 | mayue.fight | 22 June 2021, 06:19:51 UTC | [FLINK-22886][state] Fix thread leak and add unit test in RocksIncrementalSnapshotStrategy | 22 June 2021, 18:24:58 UTC |
3d93d67 | Ingo Bürk | 11 June 2021, 10:43:09 UTC | [FLINK-22788][table-planner-blink] Support equalisers for many fields When working with hundreds of fields, equalisers can fail to compile because the method body grows beyond 64kb. With this change, instead of generating all code into one method, we generate a dedicated method per field and then call all of those methods. This doesn't entirely remove the problem, but supports roughly a factor of 10 more fields and is currently deemed sufficient. This closes #16213. | 22 June 2021, 17:15:43 UTC |
4b4b614 | Michal Ciesielczyk | 26 May 2021, 11:43:59 UTC | [backport-1.12][FLINK-22698][connectors/rabbitmq] Add deliveryTimeout to RabbitMQ source This change enables setting the message delivery timeout in the RabbitMQ queueing consumer, allowing to properly stop the job in cases when no new message is available on the queue. Changes: - Add the ability to setDeliveryTimeout on the RMQConnectionConfig and its builder - Change default message delivery timeout in the RMQSource to 30 seconds (previously there was no timeout) - Extend RabbitMQ source unit tests | 22 June 2021, 11:32:58 UTC |
1e2a1ec | Xintong Song | 10 June 2021, 05:19:08 UTC | [FLINK-22312][yarn][test] Fix test instabilities due to expected heartbeat exceptions in log This closes #16232 | 22 June 2021, 09:59:53 UTC |
38bf63b | jinxing64 | 18 June 2021, 07:34:49 UTC | [FLINK-23030][network] PartitionRequestClientFactory#createPartitionRequestClient should throw when network failure | 21 June 2021, 15:11:32 UTC |
710b5f6 | Yun Gao | 18 June 2021, 07:35:24 UTC | [FLINK-21952] Make all the "Connection reset by peer" exception wrapped as RemoteTransportException | 21 June 2021, 07:26:58 UTC |
2b26df9 | Chesnay Schepler | 21 June 2021, 07:01:37 UTC | [FLINK-22987][ci] Fix scala suffix check * [FLINK-22987][ci] Fix scala suffix check * temporarily disable avro registry * Add basic precondition that at least 1 clean&infected module must be found * Invert exit handling | 21 June 2021, 07:06:59 UTC |
68961a7 | Emre Kartoglu | 17 June 2021, 16:12:03 UTC | [FLINK-23009][kinesis] Upgrade Guava for flink connector kinesis Also make flink-sql-connector-kinesis use the Guava library coming transitively from the connector-kinesis dependency. | 18 June 2021, 14:10:36 UTC |
949db6f | Guokuai Huang | 16 June 2021, 08:01:53 UTC | [FLINK-22946][runtime] Recycle floating buffer outside the lock to avoid deadlock | 18 June 2021, 12:52:13 UTC |
654e2a6 | Rui Li | 07 June 2021, 13:07:40 UTC | [FLINK-22890][hive] HiveTestUtils should create partition after the data file is ready This closes #16099 | 16 June 2021, 06:40:16 UTC |
7b4d6b8 | CodeCooker17 | 11 June 2021, 06:32:17 UTC | [FLINK-22963][doc] correct the description of taskmanager.memory.task.heap.size. This closes #16152 | 15 June 2021, 07:25:29 UTC |
faf7cc4 | TsReaper | 11 June 2021, 10:19:37 UTC | [FLINK-22015][table-planner-blink] Exclude IS NULL from SEARCH operators (#16140) | 11 June 2021, 10:19:37 UTC |
11c2030 | Fabian Paul | 11 June 2021, 10:07:35 UTC | [FLINK-22908][tests] Wait until job has started before testing shutdown | 11 June 2021, 10:07:35 UTC |
e9e8d1c | Chesnay Schepler | 23 April 2021, 17:25:54 UTC | [FLINK-22419][coordination][tests] Rework RpcEndpoint delay tests This closes #16136 | 11 June 2021, 07:25:39 UTC |
aa3cd4b | Chesnay Schepler | 10 June 2021, 14:48:10 UTC | [FLINK-22952][azure] Use Ruby 2.5 | 10 June 2021, 14:48:10 UTC |
6e86ccb | Ankush Khanna | 23 May 2021, 15:41:39 UTC | Adding gcs documentation. Connecting flink to gcs. | 09 June 2021, 18:05:23 UTC |
e794574 | Piotr Nowojski | 01 June 2021, 13:16:07 UTC | [FLINK-22833][metrics] Set checkpointStartDelay in checkpoint metrics for SourceTasks This commit is fixing a bug of checkpointStartDelay being not set in CheckpointMetrics of SourceStreamTask and SourceOperatorStreamTask. | 09 June 2021, 13:33:30 UTC |
30637ef | Piotr Nowojski | 02 June 2021, 08:06:08 UTC | [hotfix][test] Deduplicate test case between SourceStreamTask and SourceOperatorStreamTask | 09 June 2021, 13:32:08 UTC |
2c0e3f6 | Piotr Nowojski | 02 June 2021, 06:56:10 UTC | [hotfix][task] Rename triggerCheckpoint to better reflect a difference compared to other similarly named methods | 09 June 2021, 13:22:39 UTC |
277965b | Chesnay Schepler | 09 June 2021, 06:11:39 UTC | [FLINK-22939][azure] Generalize JDK switch | 09 June 2021, 09:26:00 UTC |
82431f1 | Till Rohrmann | 08 June 2021, 10:52:57 UTC | [hotfix] Replace deprecated import org.junit.Assert.assertThat with org.hamcrest.MatcherAssert.assertThat | 08 June 2021, 15:57:29 UTC |
4c05456 | Till Rohrmann | 08 June 2021, 10:52:08 UTC | [FLINK-22496][tests] Include log output from started process in ClusterEntrypointTest.testCloseAsyncShouldBeExecutedInShutdownHook This closes #16111. | 08 June 2021, 15:57:29 UTC |
b7b324d | Till Rohrmann | 08 June 2021, 10:42:18 UTC | [FLINK-22496][tests] Harden ClusterEntrypointTest.testCloseAsyncShouldBeExecutedInShutdownHook Hardens the ClusterEntrypointTest.testCloseAsyncShouldBeExecutedInShutdownHook by increasing the timeout from 3s to 10s. | 08 June 2021, 15:57:28 UTC |
9831b46 | Robert Metzger | 02 June 2021, 13:10:22 UTC | [FLINK-22856][Azure] Upgrade to ubuntu-20.04 | 08 June 2021, 10:00:06 UTC |
1575e53 | Dawid Wysakowicz | 28 May 2021, 09:07:51 UTC | [FLINK-22815][checkpointing] Remove SubtaskStateMapper.DISCARD_EXTRA_STATE which does not work The mapping implemented via DISCARD_EXTRA_STATE is not supported on the network level. At the same time there is no use for that mapping. It's better to remove the mapper for now, so that it is not used by mistake. | 08 June 2021, 06:30:40 UTC |
e6ab966 | Arvid Heise | 19 March 2021, 18:54:27 UTC | [FLINK-21945][streaming] Add the possibility to force certain exchanges to be aligned. Introduces the AlignmentType to CheckpointOptions that includes a FORCED_ALIGNED. This barrier acts as an aligned barrier for certain exchanges. The flag is then later used to restore the unaligned characteristic in SubtaskCheckpointCoordinatorImpl for downstream exchanges. | 08 June 2021, 06:30:40 UTC |
bdf4399 | Dawid Wysakowicz | 28 May 2021, 07:39:09 UTC | [FLINK-22815][checkpointing] Disable unaligned checkpoints for broadcast partitioning Broadcast partitioning can not work with unaligned checkpointing. There is no guarantees that records are consumed at the same rate in all channels. This can result in some tasks applying state changes corresponding to a certain broadcasted event while others don't. In turn upon restore it may lead to an unconsistent state. | 08 June 2021, 06:30:40 UTC |
68f1812 | Yangze Guo | 17 May 2021, 10:02:18 UTC | [FLINK-22683][runtime] Fix the null or incorrect value of total Flink/process memory in creating TaskExecutorMemoryConfiguration This closes #16067 | 04 June 2021, 06:20:05 UTC |
08bf3e6 | Yangze Guo | 05 January 2021, 09:15:18 UTC | [hotfix][coordination] Adjust TaskExecutor resource to reasonably large value for local execution This fix is to avoid arithmetic overflow in calculating the resource overview of the minicluster. | 04 June 2021, 06:20:04 UTC |
31f4099 | Fabian Paul | 01 June 2021, 13:00:16 UTC | [FLINK-22820] Allow storing SUSPENDED jobs in FileArchivedExecutionGraphStore | 03 June 2021, 07:29:22 UTC |
9f1de31 | Fabian Paul | 02 June 2021, 14:00:51 UTC | [FLINK-22756] Ignore leadership future termination during testing This changes hardens the DispatcherTest#testJobStatusIsShownDuringTermination by ignoring the JobMaster shutdown exception when querying for the expected job status. | 02 June 2021, 14:00:51 UTC |
aa590c8 | Piotr Nowojski | 31 May 2021, 15:37:55 UTC | [FLINK-22814][metrics] Calculate checkpointStartDelay for FLIP-27 sources | 01 June 2021, 12:34:20 UTC |
d215329 | Timo Walther | 25 May 2021, 13:00:51 UTC | [FLINK-22774][sql-connector-kinesis] Update Kinesis SQL connector's Guava to 27.0-jre Avoids that security tools complain about an outdated Guava version in the Kinesis SQL connector. This closes #16005. | 26 May 2021, 20:31:56 UTC |
b0b7dfc | Arvid Heise | 25 May 2021, 16:21:49 UTC | [FLINK-22613][tests] Fix FlinkKinesisITCase.testStopWithSavepoint. The test relies on a few elements remaining pending after stop-with-savepoint, however that can only be guaranteed heuristically: We cannot block task thread or else the respective savepoint will not succeed but we also cannot add infinite input as it's an IT test against Kinesis. Here, the fix is to add much more elements to Kinesis stream. An optimized sendMessage on the test client ensures timely setup. | 26 May 2021, 14:50:41 UTC |
cc885c4 | Timo Walther | 21 May 2021, 12:55:39 UTC | [FLINK-22747] Upgrade to commons-io 2.8.0 This closes #15989. | 25 May 2021, 16:44:28 UTC |
a58b516 | Arvid Heise | 21 May 2021, 14:51:02 UTC | Update japicmp configuration for 1.12.4 | 21 May 2021, 14:51:02 UTC |
5052dc3 | Fabian Paul | 12 May 2021, 11:50:22 UTC | [FLINK-22434] Retry in clusterclient if job state is unknown Since the dispatcher now exposed the SUSPENDED job state it needs to be handled in the ClusterClient. We decided to not expose this state because it would be an unexpected change for users relying on the old behavior. In summary, the RestClusterClient will retry to receive a different state for jobs in SUSPENDED state. This closes #15973. | 21 May 2021, 12:49:01 UTC |
083fe40 | Fabian Paul | 26 April 2021, 12:42:44 UTC | [FLINK-22434] Store suspended execution graphs on termination to keep them accessible Before this change as soon as a cluster termination was initiated every job was suspended and not accessible anymore until recovery. By storing the execution graph they will appear as SUSPENDED they are now accessible until the cluster is fully shutdown. | 21 May 2021, 12:49:00 UTC |
c35df7c | Dawid Wysakowicz | 19 May 2021, 07:21:11 UTC | [FLINK-22708][config] Propagate savepoint settings from StreamExecutionEnvironment to StreamGraph | 21 May 2021, 09:02:51 UTC |
1e5d808 | Till Rohrmann | 20 May 2021, 08:32:57 UTC | [FLINK-22721][ha] Add default implementation for HighAvailabilityServices.cleanupJobData In order to not break the HighAvailabilityServices interface, this commit adds a default implementation for the HighAavilabilityServices.cleanupJobData method which does nothing. Users of this interface are recommended to override this method and add a specific implementation in order to clean up job specific HA services after a job completion. This closes #15971. | 20 May 2021, 15:53:14 UTC |
7b9351e | Till Rohrmann | 19 May 2021, 08:27:46 UTC | [FLINK-22704][tests] Harden ZooKeeperHaServicesTest.testCleanupJobData This commit hardens the ZooKeeperHaServicesTest.testCleanupJobData by making sure that all zNodes have been created before checking for their existence. This is done by waiting for the leader information for the ResourceManager and the JobMaster. This closes #15956. | 20 May 2021, 08:17:47 UTC |
e09c919 | Roman Khachatryan | 15 April 2021, 11:05:21 UTC | [FLINK-21329][tests] Increase timeout and delay in local test_local_recovery_and_scheduling.sh Back-pressure is likely to occur without injection delay at the source which may lead to delaying checkpoint triggering at the sources which may lead to a timeout. To prevent this, test timeout is increased from 10m to 15m and injection delay of 100ms is added. | 19 May 2021, 12:27:58 UTC |
7e6fb13 | Roman Khachatryan | 15 April 2021, 10:58:14 UTC | [hotfix][runtime] Log checkpoint processing delay if above threshold | 19 May 2021, 12:27:58 UTC |
a153976 | Roman Khachatryan | 05 May 2021, 16:59:14 UTC | [FLINK-22502][checkpointing] Don't tolerate checkpoint retrieval failures on recovery Ignoring such failures and running with an incomplete set of checkpoints can lead to consistency violation. Instead, transient failures should be mitigated by automatic job restart. | 18 May 2021, 20:16:40 UTC |
ee34945 | Chesnay Schepler | 17 February 2021, 09:46:35 UTC | [hotfix][tests] Move test checkpoint component classes to upper level | 18 May 2021, 20:16:40 UTC |
6b53f8b | Matthias Pohl | 04 May 2021, 13:32:36 UTC | [FLINK-22494][runtime] Adds PossibleInconsistentStateException handling to CheckpointCoordinator | 18 May 2021, 09:07:27 UTC |
5aec1bb | Matthias Pohl | 04 May 2021, 15:37:35 UTC | [FLINK-22494][runtime] Refactors CheckpointsCleaner to handle also discardOnFailedStoring | 18 May 2021, 09:07:27 UTC |
b81887e | Matthias Pohl | 04 May 2021, 07:41:38 UTC | [FLINK-22494][ha] Introduces PossibleInconsistentState to StateHandleStore | 18 May 2021, 09:07:27 UTC |
95bd043 | Matthias Pohl | 04 May 2021, 11:08:07 UTC | [FLINK-22494][ha] Refactors TestingLongStateHandleHelper to operate on references The previous implementation stored the state in the StateHandle. This causes problems when deserializing the state creating a new instance that does not point to the actual state but is a copy of this state. This refactoring introduces LongStateHandle handling the actual state and LongRetrievableStateHandle referencing this handle. | 18 May 2021, 09:07:27 UTC |
a53a1f3 | Matthias Pohl | 03 May 2021, 07:40:12 UTC | [FLINK-22494][kubernetes] Introduces PossibleInconsistentStateException We experienced cases where the ConfigMap was updated but the corresponding HTTP request failed due to connectivity issues. PossibleInconsistentStateException is used to reflect cases where it's not clear whether the data was actually written or not. | 18 May 2021, 09:07:26 UTC |
814fc65 | Yi Tang | 12 April 2021, 04:59:40 UTC | [FLINK-20695][ha] Clean ha data for job if globally terminated At the moment Flink only cleans up the ha data (e.g. K8s ConfigMaps, or Zookeeper nodes) while shutting down the cluster. This is not enough for a long running session cluster to which you submit multiple jobs. In this commit, we clean up the data for the particular job if it reaches a globally terminal state. This closes #15910. | 18 May 2021, 07:20:23 UTC |
60093e4 | SteNicholas | 13 May 2021, 12:25:53 UTC | [FLINK-22592][runtime] numBuffersInLocal is always zero when using unaligned checkpoints This closes #15915 | 17 May 2021, 03:16:19 UTC |
9541b4c | Yuan Mei | 10 May 2021, 05:46:43 UTC | [FLINK-21469][runtime] Implement advanceToEndOfEventTime for MultipleInputStreamTask For stop with savepoint, StreamTask#advanceToEndOfEventTime() is called (in source tasks) to advance to the max watermark. This PR implments advanceToEndOfEventTime for MultipleInputStreamTask chained sources. | 12 May 2021, 12:20:04 UTC |
531158f | wangyang0918 | 10 November 2020, 05:58:14 UTC | [FLINK-17857][test] Make K8s e2e tests could run on Mac This closes #15890. | 11 May 2021, 12:11:20 UTC |
8c7a1c8 | wangyang0918 | 07 May 2021, 08:22:24 UTC | [hotfix][e2e] Output and collect the logs for Kubernetes IT cases | 11 May 2021, 12:11:18 UTC |
9e5b0d4 | aidenma | 27 April 2021, 14:30:57 UTC | [FLINK-22408][sql-parser] Fix SqlDropPartitions unparse Error This closes #15788 | 11 May 2021, 10:06:34 UTC |
2b1d475 | Matthias Pohl | 07 May 2021, 12:53:07 UTC | [FLINK-22566][test] Adds log extraction for the worker nodes We struggled to get the logs of the node manager which made it hard to investigate FLINK-22566 where there was a lag between setting up the YARN containers and starting the TaskExecutor. Hopefully, the nodemanager logs located on the worker nodes will help next time to investigate something like that. | 11 May 2021, 08:23:41 UTC |
cd425a2 | Matthias Pohl | 07 May 2021, 11:42:44 UTC | [hotfix][test] Adds -e flag to interpret newline in the right way | 11 May 2021, 08:23:41 UTC |
ca9d2cb | Matthias Pohl | 10 May 2021, 10:22:05 UTC | Revert "[FLINK-22566][e2e] Disable Running Kerberized YARN application on Docker test (default input) fails with no resources" This reverts commit 59f764ba504df53e5bd6226c6d8e668f1f964ba3. | 11 May 2021, 08:23:41 UTC |
4fd1344 | Roman Khachatryan | 27 April 2021, 22:02:33 UTC | [FLINK-21181][runtime] Wait for Invokable cancellation before releasing network resources | 10 May 2021, 12:14:10 UTC |
daaa01d | Till Rohrmann | 07 May 2021, 15:09:27 UTC | [FLINK-22597] Make JobMaster restartable The problem is that the DefaultLeaderRetrievalService for the ResourceManager is reused by the JobMaster. Since the DefaultLeaderRetrievalService is only usable once, this means that the JobMaster can only be started once even though the JobManagerRunnerImpl restarts it in case of a regained leadership. This closes #15854. | 10 May 2021, 10:17:55 UTC |
ed9965c | Till Rohrmann | 06 May 2021, 13:29:01 UTC | [FLINK-22577][tests] Harden KubernetesLeaderElectionAndRetrievalITCase This commit introduces closing logic to the TestingLeaderElectionEventHandler which would otherwise forward calls after the KubernetesLeaderElectionDriver is closed. This closes #15857. | 10 May 2021, 10:08:38 UTC |
87555d8 | Chesnay Schepler | 04 May 2021, 08:41:46 UTC | [FLINK-22555][build][python] Exclude leftover jboss files | 09 May 2021, 18:23:27 UTC |
b377569 | Chesnay Schepler | 05 May 2021, 11:18:11 UTC | [FLINK-22560][build] Add dedicated name to flink-dist shade-plugin execution | 09 May 2021, 18:23:27 UTC |