[
https://issues.apache.org/jira/browse/FLINK-37887?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18003739#comment-18003739
]
Nishita Pattanayak commented on FLINK-37887:
--------------------------------------------
[~imilavec] Is this config set to false in your case:
`kubernetes.operator.dynamic.config.enabled`
[https://nightlies.apache.org/flink/flink-kubernetes-operator-docs-release-1.10/docs/operations/configuration/#dynamic-operator-configuration]
Ideally with this true, Operator starts to reconcile CRDs dynamically as they
are added in the {{kubernetes.operator.watched.namespaces}} list. If false, the
pod needs to restart to take these configurations into account.
> Watched namespaces not picked up automatically by the operator
> --------------------------------------------------------------
>
> Key: FLINK-37887
> URL: https://issues.apache.org/jira/browse/FLINK-37887
> Project: Flink
> Issue Type: Bug
> Components: Kubernetes Operator
> Affects Versions: 1.10.0
> Environment: K8s: v1.30.11-eks-4096722
> Flink Operator: 1.10.0
> Reporter: Igor Milavec
> Priority: Major
>
> We use flink-kubernetes-operator Helm chart with watchNamespaces defined.
> Upon deployment the Flink Operator seems to pick up the new values and prints
> in the log:
> {noformat}
> o.a.f.c.GlobalConfiguration [INFO ] Loading configuration property:
> old-namespace-1, old-namespace-2, ..., new-namespace
> {noformat}
> However, the operator does not react to FlinkDeployment being created in the
> new namespace until we restart the flink-kubernetes-operator Deployment.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
