Zhijie Shen commented on YARN-3125:

Li, thanks for working on the patch. Looking at the test code of 
TestDistributedShell. MiniYarnCluster will be started for each individual test 
case. Therefore, we can potentially avoid conflict by configuration, and don't 
need to hard code the service address. For those test cases about v1 timeline 
service, you set enableAHS = true, while for the test cases about v2, you add 
the aux service configuration. In this way, either v1 or v2 timeline service 
will be set up, but not both. To distinguish the different test cases, you can 
try the following thing:
@Rule public TestName name = new TestName();
Using the test name to switching the setup of MiniYarnCluster in setup().

Another minor issue. Instead of using
  private static final String TIMELINE_AUX_SERVICE_CLASS =
          + ".PerNodeAggregatorServer";
You can use {{PerNodeAggregatorServer.class.getName()}} directly.

> [Event producers] Change distributed shell to use new timeline service
> ----------------------------------------------------------------------
>                 Key: YARN-3125
>                 URL: https://issues.apache.org/jira/browse/YARN-3125
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: timelineserver
>            Reporter: Zhijie Shen
>            Assignee: Junping Du
>         Attachments: YARN-3125.patch, YARN-3125_UT-022615.patch, 
> YARN-3125v2.patch, YARN-3125v3.patch
> We can start with changing distributed shell to use new timeline service once 
> the framework is completed, in which way we can quickly verify the next gen 
> is working fine end-to-end.

This message was sent by Atlassian JIRA

Reply via email to