[ 
https://issues.apache.org/jira/browse/FLINK-20797?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17256255#comment-17256255
 ] 

Yang Wang commented on FLINK-20797:
-----------------------------------

Mounting a PVC to replace the distributed storage in HA should work. Please 
make sure that you are mounting a same multiple-read-write PVC for all the 
JobManager/TaskManager pods. And it just feels like a distributed storage.

> can flink on k8s use pv using NFS and pvc as the hight avalibility storagedir
> -----------------------------------------------------------------------------
>
>                 Key: FLINK-20797
>                 URL: https://issues.apache.org/jira/browse/FLINK-20797
>             Project: Flink
>          Issue Type: New Feature
>          Components: Client / Job Submission
>         Environment: FLINK 1.12.0
>  
>            Reporter: hayden zhou
>            Priority: Major
>
> I want to deploy Flink on k8s with HA mode, and I don't want to deploy the 
> HDFS cluster, and I have an NFS so that I am created a PV that use NFS as the 
> backend storage, and I created a PVC for deployment mount.
> this is my FLINK configMap
> ```
> kubernetes.cluster-id: mta-flink
>  high-availability: org.apache.flink.kubernetes.highavailability. 
> KubernetesHaServicesFactory
>  high-availability.storageDir: file:///opt/flink/nfs/ha
> ```
> and this is my jobmanager yaml file:
> ```
> volumeMounts:
>  - name: flink-config-volume
>  mountPath: /opt/flink/conf
>  - name: flink-nfs
>  mountPath: /opt/flink/nfs
>  securityContext:
>  runAsUser: 9999 # refers to user _flink_ from official flink image, change 
> if necessary
>  #fsGroup: 9999
>  volumes:
>  - name: flink-config-volume
>  configMap:
>  name: mta-flink-config
>  items:
>  - key: flink-conf.yaml
>  path: flink-conf.yaml
>  - key: log4j-console.properties
>  path: log4j-console.properties
>  - name: flink-nfs
>  persistentVolumeClaim:
>  claimName: mta-flink-nfs-pvc
> ```
> It can be deployed successfully, but if I browser the jobmanager:8081 
> website, I get the result below:
> ```
> {"errors": ["Service temporarily unavailable due to an ongoing leader 
> election. Please refresh."]}
> ```
>  
> is the PVC can be used as `high-availability.storageDir`?  if it's can be 
> used, how can I fix this error?



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

Reply via email to