[
https://issues.apache.org/jira/browse/YARN-4807?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15192771#comment-15192771
]
Sunil G commented on YARN-4807:
-------------------------------
Yes [~kasha]. Many of the waiteState has cumulative timeout value hardcoded and
also per-round waitSecs also is hardcoded differently from method to method. So
total no: of rounds to wait for resource allocations were found by
(timeout/per-round-wait) value. As I see its, no: of times we wait OR total
number of rounds also can contribute little bit to the duration of the tests.
So if we are starting with 10ms, we can define how many rounds are also needed
for a minimum try. With few dry-runs, a good figure could be calculated.
bq.I have seen test cases where there is no reference to an RM or MockRM.
Yes, there are places even from MokRM we use this. SO as you suggested, its
better to focus on improving test time first and later some cleanup can be
tried.
> MockAM#waitForState sleep duration is too long
> ----------------------------------------------
>
> Key: YARN-4807
> URL: https://issues.apache.org/jira/browse/YARN-4807
> Project: Hadoop YARN
> Issue Type: Sub-task
> Affects Versions: 2.8.0
> Reporter: Karthik Kambatla
> Assignee: Yufei Gu
> Labels: newbie
>
> MockAM#waitForState sleep duration (500 ms) is too long. Also, there is
> significant duplication with MockRM#waitForState.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
