[ https://issues.apache.org/jira/browse/FLINK-13695?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Flink Jira Bot updated FLINK-13695: ----------------------------------- Priority: Major (was: Critical) > Integrate checkpoint notifications into StreamTask's lifecycle > -------------------------------------------------------------- > > Key: FLINK-13695 > URL: https://issues.apache.org/jira/browse/FLINK-13695 > Project: Flink > Issue Type: Bug > Components: Runtime / Task > Affects Versions: 1.9.0, 1.10.0 > Reporter: Till Rohrmann > Priority: Major > Labels: auto-deprioritized-critical > > The {{StreamTask's}} asynchronous checkpoint notifications are decoupled from > the {{StreamTask's}} lifecycle. Consequently, it can happen that a > {{StreamTask}} is terminating/cancelling and still sends asynchronous > checkpoint notifications (e.g. acknowledge/decline checkpoint notifications). > This is problematic because a cancelling/terminating {{StreamTask}} might > cause an asynchronous checkpoint to fail which is expected and should not be > reported to the {{JobMaster}}. > Hence, the checkpoint notifications should be coupled with the > {{StreamTask's}} lifecycle (e.g. notifications must be sent from the > {{StreamTask's}} main thread) and if not valid, then they need to be filtered > out/suppressed. -- This message was sent by Atlassian Jira (v8.3.4#803005)