[ https://issues.apache.org/jira/browse/YARN-8085?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16418494#comment-16418494 ] 

Rohith Sharma K S commented on YARN-8085:
-----------------------------------------

thanks [~Tao Yang] for the patch! It is a blocker for 3.1.0.

Regarding the patch, we have clear separation of alwaysOnService vs activeServiceContext. ResourceProfilesManager should be part of active service context because activeServiceContext context is refreshed  for every transition. I see that profiles are loaded during service initialization. If it get added into serviceContext, then profiles are NOT refreshed which get stale profile information. cc :/ [~sunilg]

> RMContext#resourceProfilesManager is lost after RM went standby then back to active
> -----------------------------------------------------------------------------------
>
>                 Key: YARN-8085
>                 URL: https://issues.apache.org/jira/browse/YARN-8085
>             Project: Hadoop YARN
>          Issue Type: Bug
>          Components: capacityscheduler
>    Affects Versions: 3.1.0
>            Reporter: Tao Yang
>            Assignee: Tao Yang
>            Priority: Major
>         Attachments: YARN-8085.001.patch, YARN-8085.002.patch
>
>
> We submited a distributed shell application after RM failover and back to active, then got NPE error in RM log:
> {noformat}
> java.lang.NullPointerException
>         at org.apache.hadoop.yarn.server.resourcemanager.ClientRMService.getResourceProfiles(ClientRMService.java:1814)
>         at org.apache.hadoop.yarn.api.impl.pb.service.ApplicationClientProtocolPBServiceImpl.getResourceProfiles(ApplicationClientProtocolPBServiceImpl.java:657)
>         at org.apache.hadoop.yarn.proto.ApplicationClientProtocol$ApplicationClientProtocolService$2.callBlockingMethod(ApplicationClientProtocol.java:617)
>         at org.apache.hadoop.ipc.ProtobufRpcEngine$Server$ProtoBufRpcInvoker.call(ProtobufRpcEngine.java:523)
>         at org.apache.hadoop.ipc.RPC$Server.call(RPC.java:991)
>         at org.apache.hadoop.ipc.Server$RpcCall.run(Server.java:869)
>         at org.apache.hadoop.ipc.Server$RpcCall.run(Server.java:815)
>         at java.security.AccessController.doPrivileged(Native Method)
>         at javax.security.auth.Subject.doAs(Subject.java:422)
>         at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1682)
>         at org.apache.hadoop.ipc.Server$Handler.run(Server.java:2675)
> {noformat}
> The cause is that currently resourceProfilesManager is not transferred to new RMContext instance in RMContext#resetRMContext. We should do this transfer to fix this error.
> {code:java}
> @@ -1488,6 +1488,10 @@ private void resetRMContext() {
>      // transfer service context to new RM service Context
>      rmContextImpl.setServiceContext(rmContext.getServiceContext());
> +    // transfer resource profiles manager
> +    rmContextImpl
> +        .setResourceProfilesManager(rmContext.getResourceProfilesManager());
> +
>      // reset dispatcher
>      Dispatcher dispatcher = setupDispatcher();
>      ((Service) dispatcher).init(this.conf);
> {code}



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscribe@hadoop.apache.org
For additional commands, e-mail: yarn-issues-help@hadoop.apache.org