[
https://issues.apache.org/jira/browse/HDFS-13746?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16576903#comment-16576903
]
Daniel Templeton commented on HDFS-13746:
-----------------------------------------
Thanks for the patch, [~smeng]. Yeah, that test is known flaky. Here are my
comments:
* Why {{Groups.LOG}}? This output has nothing to do with that class. I'd just
create my own logger for the test class. For one of many, many examples, see
{{TestImmutableSnapshot}}.
* On your sleep timing, I don't see the need for the initial 1000ms *
{{groupRefreshTimeoutSec}} sleep. I also don't see that the max number of
trials is governed by the {{groupRefreshTimeoutSec}}. You'd be better off
dropping the first sleep and making the max number of trials be a factor of
{{groupRefreshTimeoutSec}}.
* Logging at WARN level should be for things that are going wrong and can be
addressed. Your log messages in this patch should be INFO or DEBUG.
* I don't think the log message inside the loop is helpful. It produces a ton
of output that makes the rest of the logs harder to read. Really what you're
trying to show if how many trials it took to pass. So say that instead when
it's done, e.g. "Cache timeout took 1250ms to take effect."
* Why did you need to write {{isSameStrList()}} instead of using
{{List.equals()}}?
> Still occasional "Should be different group" failure in
> TestRefreshUserMappings#testGroupMappingRefresh
> -------------------------------------------------------------------------------------------------------
>
> Key: HDFS-13746
> URL: https://issues.apache.org/jira/browse/HDFS-13746
> Project: Hadoop HDFS
> Issue Type: Bug
> Affects Versions: 3.0.0
> Reporter: Siyao Meng
> Assignee: Siyao Meng
> Priority: Major
> Attachments: HDFS-13746.001.patch, HDFS-13746.002.patch
>
>
> In https://issues.apache.org/jira/browse/HDFS-13723, increasing the amount of
> time in sleep() helps but the problem still appears, which is annoying.
>
> Solution:
> Use a loop to allow the test case to fail 100(MAX_RETRIES) times before
> declaring failure. Wait 50 ms between each retry.
>
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
