[ https://issues.apache.org/jira/browse/FLINK-5218?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15712711#comment-15712711 ]
ASF GitHub Bot commented on FLINK-5218: --------------------------------------- Github user StephanEwen commented on the issue: https://github.com/apache/flink/pull/2920 @StefanRRichter There is one change of semantics that would be good to get your input on: A checkpoint stream to which a `byte[0]` array was written is now actually empty and returns a `null` state handle in the same way as if nothing was ever written. Before this change, it would have created a state of zero bytes. > Eagerly close checkpoint streams on cancellation > ------------------------------------------------ > > Key: FLINK-5218 > URL: https://issues.apache.org/jira/browse/FLINK-5218 > Project: Flink > Issue Type: Bug > Components: State Backends, Checkpointing > Affects Versions: 1.1.3 > Reporter: Stephan Ewen > Assignee: Stephan Ewen > Priority: Critical > Fix For: 1.2.0, 1.1.4 > > > Some output streams perform blocking operations that cannot be properly > interrupted. This causes cancellations to take very long when happening > concurrently to large synchronous state snapshot operations. > Closing the streams concurrently helps to abort these blocking operations. > This might already be fixed in 1.2 by the {{CloseableRegistry}}. -- This message was sent by Atlassian JIRA (v6.3.4#6332)