hadoop-yarn-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Bikas Saha (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (YARN-513) Create common proxy client for communicating with RM
Date Tue, 30 Apr 2013 22:04:18 GMT

    [ https://issues.apache.org/jira/browse/YARN-513?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13646048#comment-13646048
] 

Bikas Saha commented on YARN-513:
---------------------------------

I like RMProxy.
Do we see this being useful to user code. If yes, then we should put it in yarn client package.


Why is this called defaultPolicy?
{code}
+  private final RetryPolicy defaultPolicy;
{code}

Can these be inferred from Protocol type? Else we cannot use this for other protocols.
{code}
+    this.rmAddress = conf.getSocketAddr(
+        YarnConfiguration.RM_RESOURCE_TRACKER_ADDRESS,
+        YarnConfiguration.DEFAULT_RM_RESOURCE_TRACKER_ADDRESS,
+        YarnConfiguration.DEFAULT_RM_RESOURCE_TRACKER_PORT);
{code}
With this the code will start looking like NameNodeProxies with its create* methods.

I looked at NNProxies class. It looks we need an almost exact version of that for RM. The
main difference being that we dont have HA yet and RM protocols run on different ports.
The neat thing about NNProxies looks like they create a transparent proxy using java.reflect.proxy
that ends up calling realImpl.invoke(method) internally when the client calls calls proxy.method().
Thus it looks better than our approach where our client explicitly call proxy.invoke("method").
Perhaps we should try to change our impl to something similar to NNProxies. What do folks
think?
                
> Create common proxy client for communicating with RM
> ----------------------------------------------------
>
>                 Key: YARN-513
>                 URL: https://issues.apache.org/jira/browse/YARN-513
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: resourcemanager
>            Reporter: Bikas Saha
>            Assignee: Xuan Gong
>         Attachments: YARN-513.1.patch, YARN-513.2.patch, YARN-513.3.patch, YARN-513.4.patch
>
>
> When the RM is restarting, the NM, AM and Clients should wait for some time for the RM
to come back up.

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Mime
View raw message