[ https://issues.apache.org/jira/browse/FLINK-5229?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15715296#comment-15715296 ]
ASF GitHub Bot commented on FLINK-5229: --------------------------------------- GitHub user tillrohrmann opened a pull request: https://github.com/apache/flink/pull/2924 [FLINK-5229] [tm] Discard completed checkpoints if a subsequent operator's checkpoint fails In case of a failure of any StreamOperator#snapshotState method, all up to this point created StreamTaskStates will be discarded. This ensures that a failing checkpoint operation of a chained operator won't leave orphaned checkpoint data behind. cc @uce, @StephanEwen. You can merge this pull request into a Git repository by running: $ git pull https://github.com/tillrohrmann/flink fixCheckpointCleanup Alternatively you can review and apply these changes as the patch at: https://github.com/apache/flink/pull/2924.patch To close this pull request, make a commit to your master/trunk branch with (at least) the following in the commit message: This closes #2924 ---- commit 3ce8cec2cb3dc0c4979ab2cbe30dc1db8d988111 Author: Till Rohrmann <trohrm...@apache.org> Date: 2016-12-02T14:33:06Z [FLINK-5229] [tm] Discard completed checkpoints if a subsequent operator's checkpoint fails In case of a failure of any StreamOperator#snapshotState method, all up to this point created StreamTaskStates are discarded. This ensures that a failing checkpoint operation of a chained operator won't leave orphaned checkpoint data behind. ---- > Cleanup StreamTaskStates if a checkpoint operation of a subsequent operator > fails > ---------------------------------------------------------------------------------- > > Key: FLINK-5229 > URL: https://issues.apache.org/jira/browse/FLINK-5229 > Project: Flink > Issue Type: Bug > Components: State Backends, Checkpointing, TaskManager > Affects Versions: 1.2.0, 1.1.3 > Reporter: Till Rohrmann > Assignee: Till Rohrmann > Fix For: 1.2.0, 1.1.4 > > > Due to chaining, a {{StreamTask}} needs to checkpoint multiple operators. If > the first operators succeed in creating a checkpoint but a subsequent > operator in the chain fails, the {{StreamTask}} has to clean up the already > completed checkpoints. Otherwise we might end up with orphaned state data. -- This message was sent by Atlassian JIRA (v6.3.4#6332)