[
https://issues.apache.org/jira/browse/FLINK-32667?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17753091#comment-17753091
]
Fang Yong edited comment on FLINK-32667 at 8/11/23 8:00 AM:
------------------------------------------------------------
[~mapohl] Thanks for your detailed explanation and I caught it. Currently Flink
will create `FailoverStrategy` for jobs according to the cluster configuration
and as I mentioned above there are only two strategies: full and region.
I think decoupling `LeaderElection` and `job-related HA data` from
`HighAvailabilityServices` is a very good solution, that's what we want for
OLAP queries.
As you mentioned in
[FLINK-31816|https://issues.apache.org/jira/browse/FLINK-31816?focusedCommentId=17741054&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-17741054]:
```
HighAvailabilityServices could get a single implementation that requires a
factory method for creating JobGraphStore, JobResultStore,
CheckpointRecoveryFactory, and BlobStore. Additionally, it would require a
LeaderElectionService (which is essentially a factory for LeaderElection
instances)
```
I think we can do it now and after that we can add a new failover strategy such
as `none` for cluster and create embedding factory. What do you think?
[~mapohl][~chesnay]
was (Author: zjureel):
[~mapohl] Thanks for your detailed explanation and I caught it. Currently Flink
will create `FailoverStrategy` for jobs according to the cluster configuration
and as I mentioned above there are only two strategies: full and region.
I think decoupling `LeaderElection` and `job-related HA data` from
`HighAvailabilityServices` is a very good solution, that's what we want for
OLAP queries.
As you mentioned in
[FLINK-31816|https://issues.apache.org/jira/browse/FLINK-31816?focusedCommentId=17741054&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-17741054]:
`HighAvailabilityServices could get a single implementation that requires a
factory method for creating JobGraphStore, JobResultStore,
CheckpointRecoveryFactory, and BlobStore. Additionally, it would require a
LeaderElectionService (which is essentially a factory for LeaderElection
instances)`
I think we can do it now and after that we can add a new failover strategy such
as `none` for cluster and create embedding factory. What do you think?
[~mapohl][~chesnay]
> Use standalone store and embedding writer for jobs with no-restart-strategy
> in session cluster
> ----------------------------------------------------------------------------------------------
>
> Key: FLINK-32667
> URL: https://issues.apache.org/jira/browse/FLINK-32667
> Project: Flink
> Issue Type: Sub-task
> Components: Runtime / Coordination
> Affects Versions: 1.18.0
> Reporter: Fang Yong
> Assignee: Fang Yong
> Priority: Major
> Labels: pull-request-available
>
> When a flink session cluster use zk or k8s high availability service, it will
> store jobs in zk or ConfigMap. When we submit flink olap jobs to the session
> cluster, they always turn off restart strategy. These jobs with
> no-restart-strategy should not be stored in zk or ConfigMap in k8s
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
