You are viewing a plain text version of this content. The canonical link for it is here.

Posted to commits@airflow.apache.org by "Ram (Jira)" <ji...@apache.org> on 2020/01/16 12:13:00 UTC

[jira] [Created] (AIRFLOW-6580) Killing or marking a task does not kill the Pod in the backend

Ram created AIRFLOW-6580:
----------------------------

             Summary: Killing or marking a task does not kill the Pod in the backend
                 Key: AIRFLOW-6580
                 URL: https://issues.apache.org/jira/browse/AIRFLOW-6580
             Project: Apache Airflow
          Issue Type: Bug
          Components: DAG, executor-kubernetes
    Affects Versions: 1.10.2
            Reporter: Ram
            Assignee: Daniel Imberman


We're using KubernetesPodOperator in Airflow 1.10.2
 The pods that we have some NodeAffinity and Tolerations in it. 
 Sometimes the pod gets stuck at a Pending state.
 *But when the task fails, the Pending pod does not kill itself.*

Related to this, when we manually fail a task, the DAG task stops running, but the Pod in the DAG does not get killed and continues running.

We have tried setting the 'is_delete_operator_pod' to True. But for some reason the Pod gets killed almost instantly when the execution starts. We have not been able to debug the issue behind this.

Does the latest version of Airflow account for this?



--
This message was sent by Atlassian Jira
(v8.3.4#803005)