[
https://issues.apache.org/jira/browse/PHOENIX-5810?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17076387#comment-17076387
]
Istvan Toth commented on PHOENIX-5810:
--------------------------------------
Looking at the docs at
https://hadoop.apache.org/docs/r2.4.1/hadoop-yarn/hadoop-yarn-common/yarn-default.xml
and
https://hadoop.apache.org/docs/current/hadoop-yarn/hadoop-yarn-site/ResourceManagerHA.html
It seems that while this code handles the non-HA case better, it is still not
ideal.
I believe that you should read *yarn.resourcemanager.webapp.address.rm-id* /
*yarn.resourcemanager.webapp.address* for this. These variables should contain
the web server address+port string that you can directly use. There is no need
for the separate port querying function either. (There is also the question of
not handling HTTP/HTTPS schema, but that may be out of scope for this ticket.)
> PhoenixMRJobSubmitter is not working on a cluster with a single yarn RM
> -----------------------------------------------------------------------
>
> Key: PHOENIX-5810
> URL: https://issues.apache.org/jira/browse/PHOENIX-5810
> Project: Phoenix
> Issue Type: Bug
> Reporter: Richard Antal
> Assignee: Richard Antal
> Priority: Major
> Attachments: PHOENIX-5810.master.v1.patch,
> PHOENIX-5810.master.v2.patch, PHOENIX-5810.master.v3.patch,
> PHOENIX-5810.master.v4.patch
>
>
> {code:java}
> Exception in thread "main"
> org.apache.zookeeper.KeeperException$NoNodeException: KeeperErrorCode =
> NoNode for /yarn-leader-election{code}
> The error happens when we want to run scheduleIndexBuilds. In
> getSubmittedYarnApps, getActiveResourceManagerHost uses zookeeper to
> determine the active Resource Manager.
> But /yarn-leader-election only exists if yarn is in HA mode.
> I think this function should work well when we have single yarn RM and read
> its address from the config.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
