[
https://issues.apache.org/jira/browse/FLINK-5943?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15988586#comment-15988586
]
ASF GitHub Bot commented on FLINK-5943:
---------------------------------------
Github user tillrohrmann commented on the issue:
https://github.com/apache/flink/pull/3730
Hi @zhangminglei, I think the logic to access the `RunningJobsRegistry`
from the `YarnApplicationMasterRunner` is wrong. The decision to clear a job or
not should be done by the `JobManagerRunner`. Therefore, the right solution
would be to remove the `clearJob` call from the `shutdown` method. I will do
this. Consequently, please close this PR and the corresponding JIRA issue.
> Unprotected access to haServices in
> YarnFlinkApplicationMasterRunner#shutdown()
> -------------------------------------------------------------------------------
>
> Key: FLINK-5943
> URL: https://issues.apache.org/jira/browse/FLINK-5943
> Project: Flink
> Issue Type: Bug
> Components: YARN
> Reporter: Ted Yu
> Assignee: mingleizhang
> Priority: Minor
>
> {code}
> protected void shutdown(ApplicationStatus status, String msg) {
> // Need to clear the job state in the HA services before shutdown
> try {
> haServices.getRunningJobsRegistry().clearJob(jobGraph.getJobID());
> }
> {code}
> The access to haServices is without lock protection.
> haServices may have been closed.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
