[
https://issues.apache.org/jira/browse/BOOKKEEPER-839?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14335202#comment-14335202
]
Rakesh R commented on BOOKKEEPER-839:
-------------------------------------
Thanks [~fpj] for sharing the logs. I have the following observation:
- {{AuditorPeriodicCheckTest#testPeriodicCheckWhenDisabled}} test is creating
100 ledgers and adding 100 entries to each ledger. Total of about 10000 ledger
entries are adding. This is costlier operation. Secondly, it kills one bookie
and start new bookie. Now the test case is expecting to detect the
under-replicated ledgers and performs re-replication of these 100 ledgers to
the new Bookie. This is again another costlier operation.
- Another observation about this test case is, {{Thread.sleep(CHECK_INTERVAL *
2000);}} is used four times, so total of 8seconds sleeping time . Probably need
to see better way by avoiding these sleeps.
In my environment it took around 30 seconds to finish the test case. Could you
try reducing the number of ledgers and the number of entries in each ledger.
Probably can try with 10 ledgers and 1 or 2 entries each. I feel this will make
the test simple.
{code}
final int numLedgers = 100;
final int numMsgs = 100;
{code}
> AuditorPeriodicCheckTest timeout
> --------------------------------
>
> Key: BOOKKEEPER-839
> URL: https://issues.apache.org/jira/browse/BOOKKEEPER-839
> Project: Bookkeeper
> Issue Type: Bug
> Reporter: Flavio Junqueira
> Attachments:
> org.apache.bookkeeper.replication.AuditorPeriodicCheckTest-output.txt.gz
>
>
> This is failing consistently for me. I'm using a ubuntu vm with Oracle jdk 8.
> Attaching a log file that I got from running it on the master branch.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
