hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Hadoop QA (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HADOOP-1870) [hbase] Parameters for TestDFSAbort too aggressive
Date Tue, 11 Sep 2007 23:31:32 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-1870?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_12526632
] 

Hadoop QA commented on HADOOP-1870:
-----------------------------------

+1

http://issues.apache.org/jira/secure/attachment/12365589/patch.txt applied and successfully
tested against trunk revision r574654.

Test results:   http://lucene.zones.apache.org:8080/hudson/job/Hadoop-Patch/734/testReport/
Console output: http://lucene.zones.apache.org:8080/hudson/job/Hadoop-Patch/734/console

> [hbase] Parameters for TestDFSAbort too aggressive
> --------------------------------------------------
>
>                 Key: HADOOP-1870
>                 URL: https://issues.apache.org/jira/browse/HADOOP-1870
>             Project: Hadoop
>          Issue Type: Bug
>          Components: contrib/hbase
>    Affects Versions: 0.15.0
>            Reporter: Jim Kellerman
>            Assignee: Jim Kellerman
>             Fix For: 0.15.0
>
>         Attachments: patch.txt, patch.txt
>
>
> The parameter settings for client retries both at the Hadoop IPC level and at the HBase
client level seem to aggressive. Even though lucene.zones.apache.org is a quad CPU machine,
it routinely displays a load average of ~8. It runs much slower than any of our test machines:
Core 2 duo MacBook Pro (running Linux or MacOs), Dual dual-core opterons (Linux),  Core 2
(Windows) and even a PentiumM (Windows).
> Consequently, tests that run just fine in the environments listed above sometimes fail
on lucene.zones. 
> It is unclear if this is due to thread scheduling differences between Solaris and Linux/Windows
or in the respective JVM's.
> To make some tests run reliably in the Hudson build environment, we sometimes have to
increase retries and timeouts, which we set aggressively so that the tests will run faster.
> If that is what it takes to prevent Hudson build failures, then so be it. The tests may
take longer to run, but if that is the tradeoff to achieve reliability, then so be it.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message