[ https://issues.apache.org/jira/browse/AIRFLOW-4715?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16852474#comment-16852474 ]
Palanieppan Muthiah commented on AIRFLOW-4715: ---------------------------------------------- Looks like this is duplicate of [https://github.com/apache/airflow/pull/4082]. So closing this. > On success callback on task does not run reliably > ------------------------------------------------- > > Key: AIRFLOW-4715 > URL: https://issues.apache.org/jira/browse/AIRFLOW-4715 > Project: Apache Airflow > Issue Type: Bug > Components: DAG > Affects Versions: 1.10.1 > Reporter: Palanieppan Muthiah > Priority: Minor > > When a task succeeds, the on_success_callback is only called after the task > state is set to success. If the success_callback runs long enough, the task > heartbeat process finds the task state is already success and kills itself, > disrupting the long running success_callback. > On the other hand, in the case of failure, the on_failure_callback is called > first before the state is set to failed. > It seems the correct behavior is that the success callback is called first, > and then state is set to success (regardless of any error in the > success_callback). -- This message was sent by Atlassian JIRA (v7.6.3#76005)