You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Andrey Zagrebin (Jira)" <ji...@apache.org> on 2019/09/05 08:57:00 UTC
[jira] [Commented] (FLINK-13895) Client does not exit when
bin/yarn-session.sh come fail
[ https://issues.apache.org/jira/browse/FLINK-13895?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16923213#comment-16923213 ]
Andrey Zagrebin commented on FLINK-13895:
-----------------------------------------
From the logs, looks like the application killing hangs because the client cannot connect to the yarn cluster RM, some networking, non-Flink issue per se.
The ConfiguredRMFailoverProxyProvider could be probably reconfigured to do limited number of reconnection retries and prevent Flink cli from hanging.
From the source code of ConfiguredRMFailoverProxyProvider.init, it looks like yarn.client.failover-retries is the option to tweak (if the default zero value probably means infinite retries). Not sure whether it makes to tweak this option in Flink for Yarn deployments by default.
[~yuwang0917@gmail.com] could you try to set [yarn.client.failover-retries or yarn.client.failover-retries-on-socket-timeouts|[https://hadoop.apache.org/docs/r2.4.1/hadoop-yarn/hadoop-yarn-common/yarn-default.xml]] to some small value to see if reconnection attempts stop and cli exits?
> Client does not exit when bin/yarn-session.sh come fail
> -------------------------------------------------------
>
> Key: FLINK-13895
> URL: https://issues.apache.org/jira/browse/FLINK-13895
> Project: Flink
> Issue Type: Improvement
> Components: Deployment / YARN
> Affects Versions: 1.9.0
> Reporter: Yu Wang
> Priority: Minor
> Labels: pull-request-available
> Attachments: client_exit.txt
>
> Time Spent: 10m
> Remaining Estimate: 0h
>
> the hadoop cluster environment java version is 1.7, flink is compiled with jdk1.8，I used bin/yarn-session.sh submit it , then client comes error and does not exit . I found yarn application which is failed , so then we should not kill the yarn application, we can stop the yarn client . attachments is operation log
--
This message was sent by Atlassian Jira
(v8.3.2#803003)