[
https://issues.apache.org/jira/browse/AIRFLOW-6062?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17091790#comment-17091790
]
ASF GitHub Bot commented on AIRFLOW-6062:
-----------------------------------------
mppetkov edited a comment on pull request #7123:
URL: https://github.com/apache/airflow/pull/7123#issuecomment-619146491
Hello, I have migrated to Airflow 1.10.10 and I am trying to use this great
feature, but unfortunately when all pods from different namespaces are moved to
status completed, then they are not picked up and deleted by the scheduler.
Correct me if I am wrong, but from what I see in the `KubernetesExecutor` on
line 368, it is still watching the main namespace only.
`for event in watcher.stream(kube_client.list_namespaced_pod,
self.namespace,`
What do you think about using `kube_client.list_pod_for_all_namespaces`
rather than `kube_client.list_namespaced_pod`?
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
> Scheduler doesn't delete worker pods from namespaces different than the
> scheduler's namespace
> ---------------------------------------------------------------------------------------------
>
> Key: AIRFLOW-6062
> URL: https://issues.apache.org/jira/browse/AIRFLOW-6062
> Project: Apache Airflow
> Issue Type: Bug
> Components: executor-kubernetes
> Affects Versions: 1.10.5
> Reporter: Mihail Petkov
> Assignee: Daniel Imberman
> Priority: Blocker
> Fix For: 1.10.10
>
>
> When you run Airflow's task instances as worker pods in different namespaces
> into a Kubernetes cluster, the scheduler can delete only the pods that are
> living in the same namespace where the scheduler lives. It's trying to delete
> all pods that are in the namespace defined in the airflow.cfg file.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
