[
https://issues.apache.org/jira/browse/HADOOP-10090?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ivan Mitic updated HADOOP-10090:
--------------------------------
Attachment: HADOOP-690.patch
Hi Chris, Luke,
I am attaching the fix for the #2 approach. The patch includes a unittest that
catches the problem.
This fix relaxes the "inconsistent result" issue, however it does not eliminate
it completely. JMX will always return complete result, but the sink might miss
some changes. The problem is that MetricsSourceAdapter must synchronize its
calls to MetricsSource#getMetrics, otherwise, we can have two concurrent
threads snapshotting the metrics, leading to one of them not getting all
metrics that had changed (if "all" is not specified, like in the sink case).
Because of the issue described in HADOOP-8050 it is not possible to just add a
lock as it can introduce a deadlock (we would also have to eliminate the system
source issue).
Let me know what you think.
> Jobtracker metrics not updated properly after execution of a mapreduce job
> --------------------------------------------------------------------------
>
> Key: HADOOP-10090
> URL: https://issues.apache.org/jira/browse/HADOOP-10090
> Project: Hadoop Common
> Issue Type: Bug
> Components: metrics
> Affects Versions: 1.2.1
> Reporter: Ivan Mitic
> Assignee: Ivan Mitic
> Attachments: HADOOP-690.patch, OneBoxRepro.png
>
>
> After executing a wordcount mapreduce sample job, jobtracker metrics are not
> updated properly. Often times the response from the jobtracker has higher
> number of job_completed than job_submitted (for example 8 jobs completed and
> 7 jobs submitted).
> Issue reported by Toma Paunovic.
--
This message was sent by Atlassian JIRA
(v6.1#6144)
