[
https://issues.apache.org/jira/browse/YARN-1165?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13761516#comment-13761516
]
Bikas Saha commented on YARN-1165:
----------------------------------
HADOOP-9933 would be a pre-requisite to restartable services but we would also
need changes in the services themselves to make them restartable in practice.
Also, what would be the behavior of other services that inherit from abstract
service but dont implement the restartable feature?
Do you want to estimate how long it will take to make these services
restartable? If thats going to be a long time, then we might have to be
tactical and use the approach in 1027-3.patch for now. Its not ideal but
practical.
The other thing we can see is to see why the tests are failing. If there is a
common issue across all tests then we may invest in fixing that. Can you please
provide an example of a test failure to explain whats happening.
> Move init() of activeServices to ResourceManager#serviceStart()
> ---------------------------------------------------------------
>
> Key: YARN-1165
> URL: https://issues.apache.org/jira/browse/YARN-1165
> Project: Hadoop YARN
> Issue Type: Sub-task
> Components: resourcemanager
> Reporter: Karthik Kambatla
> Assignee: Karthik Kambatla
> Attachments: test-failures.pdf
>
>
> Background:
> # YARN-1098 separates out RM services into Always-On and Active services, but
> doesn't change the behavior in any way.
> # For YARN-1027, we would want to create, initialize, and start
> RMActiveServices in the scope of RM#serviceStart(). This requires updating
> test cases that check for certain behavior post RM#serviceInit() - otherwise,
> most of these tests NPE.
> Creating a JIRA different from YARN-1027 to address all these test cases.
--
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
