[
https://issues.apache.org/jira/browse/FLINK-37391?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18003740#comment-18003740
]
Nishita Pattanayak commented on FLINK-37391:
--------------------------------------------
[https://github.com/apache/flink-kubernetes-operator/blob/8812c78cd6a2c0ad1b672ca08a8b880bd890ae8b/flink-kubernetes-operator/src/main/java/org/apache/flink/kubernetes/operator/config/FlinkConfigManager.java#L144]
There is a flag kubernetes.operator.dynamic.config.enabled which enables update
of the watched namespaces on the fly by ensuring a watcher tracks the changes
to config in a fixed interval
> Change watchNamespace config in operator is not respected while running
> -----------------------------------------------------------------------
>
> Key: FLINK-37391
> URL: https://issues.apache.org/jira/browse/FLINK-37391
> Project: Flink
> Issue Type: Bug
> Components: Kubernetes Operator
> Affects Versions: 1.10.0, 1.19.0
> Reporter: Fabian Paul
> Priority: Major
>
> The Flink operator allows the limiting of the watched namespaces by setting
> the `watchNamespaces` config during helm installation and upgrades.
>
> Currently, the configuration is only loaded into operator during the initial
> start
> [https://github.com/apache/flink-kubernetes-operator/blob/9eb3c385b90a5a2f08376720f3204d1784981a0c/flink-kubernetes-operator/src/main/java/org/apache/flink/kubernetes/operator/FlinkOperator.java#L247]
> leading to that updating the watched namespaces doesn't inform the java
> process and only create the necessary k8s resources, e.g., service account
> via the helm upgrade.
>
> One idea could be that changing watchNamespaces triggers a restart of the
> operator pod and, hence reinitializes the watch behavior.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
