[jira] [Updated] (KAFKA-16889) Kafka pods are not gracefully terminated

["Sreelekshmi Geetha (Jira)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/KAFKA-16889?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Sreelekshmi Geetha updated KAFKA-16889:
---------------------------------------
    Description: 
The SIGTERM signal is not properly handled by Kafka container and the pod is forcibly killed by kubernetes with exit code 137.

When we try kill -15 1 on the Kafka container, its exits with code 137, but should be 143. The kafka process is actually shutdown properly and unregistered form the cluster.

The issue is that after it's unregistered from the cluster the whole pod is not restarted. You have to wait for the liveness probe to fail which will as a consequence restart kafka container.

  was:
The SIGTERM signal is not properly handled by Kafka container and the pod is forcibly killed by kubernetes with exit code 137.

 

When we try kill -15 1 on the Kafka container, its exits with code 137, but should be 143.


> Kafka pods are not gracefully terminated
> ----------------------------------------
>
>                 Key: KAFKA-16889
>                 URL: https://issues.apache.org/jira/browse/KAFKA-16889
>             Project: Kafka
>          Issue Type: Bug
>    Affects Versions: 3.5.1
>            Reporter: Sreelekshmi Geetha
>            Priority: Major
>
> The SIGTERM signal is not properly handled by Kafka container and the pod is forcibly killed by kubernetes with exit code 137.
> When we try kill -15 1 on the Kafka container, its exits with code 137, but should be 143. The kafka process is actually shutdown properly and unregistered form the cluster.
> The issue is that after it's unregistered from the cluster the whole pod is not restarted. You have to wait for the liveness probe to fail which will as a consequence restart kafka container.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)