[
https://issues.apache.org/jira/browse/AIRFLOW-3485?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
曾政 updated AIRFLOW-3485:
------------------------
Priority: Major (was: Critical)
> task rerun after celery worker restart
> --------------------------------------
>
> Key: AIRFLOW-3485
> URL: https://issues.apache.org/jira/browse/AIRFLOW-3485
> Project: Apache Airflow
> Issue Type: Bug
> Components: celery, cli
> Affects Versions: 1.10.1
> Environment: centos7
> Reporter: 曾政
> Assignee: 曾政
> Priority: Major
> Original Estimate: 12h
> Remaining Estimate: 12h
>
> I find a bug.
> I'm using redis as broker，and using CeleryExecutor as executor。
> step 1:
> force rerun a task that already in success state.
> for example: airflow run dag_id task_name execution_date -A -i --force
> step 2:
> before the task end kill the celery worker (or maybe the worker crash or
> else); and the task is still in running.
> step 3:
> waiting for the task end.
> step 4:
> restart worker and you will find the task will rerun.
>
> I research the problem and find the reason . when killing worker ,the worker
> restore the command to the queue .when the worker restart it read the same
> command and execute it.
> but the normal scheduled task will check its state and ignore the command.
> I find a method to fix it.
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
