tillrohrmann commented on a change in pull request #18808:
URL: https://github.com/apache/flink/pull/18808#discussion_r808859824
##########
File path:
docs/content/docs/deployment/resource-providers/standalone/kubernetes.md
##########
@@ -246,6 +246,17 @@ To use Reactive Mode on Kubernetes, follow the same steps
as for [deploying a jo
Once you have deployed the *Application Cluster*, you can scale your job up or
down by changing the replica count in the `flink-taskmanager` deployment.
+### Enabling Local Recovery Across Pod Restarts
+
+In order to speed up recoveries in case of pod failures, you can leverage
Flink's [working directory]({{< ref
"docs/deployment/resource-providers/standalone/working_directory" >}}) feature
together with local recovery.
+If the working directory is configured to reside on a persistent volume that
gets remounted to a restarted TaskManager pod, then Flink is able to recover
state locally.
+With the
[StatefulSet](https://kubernetes.io/docs/concepts/workloads/controllers/statefulset/),
Kubernetes gives you the exact tool you need to map a pod to a persistent
volume.
+
+So instead of deploying the TaskManagers as a Deployment, you need to
configure a StatefulSet for the TaskManagers.
+The StatefulSet allows to configure a volume claim template that you use to
mount persistent volumes to the TaskManagers.
Review comment:
We also need a mapping between pod and persistent volume. This is not
possible with a deployment alone because then every pod would get the same
claim assigned.
--
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.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
