[ 
https://issues.apache.org/jira/browse/HADOOP-10090?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13824041#comment-13824041
 ] 

Luke Lu commented on HADOOP-10090:
----------------------------------

I now recalled some hesitation of extra lock on source:

* It could adversely affect the application performance buy holding the source 
lock while doing a snapshot. Currently source have a choice on whether and how 
the snapshot should be synchronized or not depending on the nature of the 
metrics involved. In many cases, source is implemented by a real object the has 
application locking logic. Holding a lock doing a potentially large snapshot 
(many metrics) _could_ increase lock contention significantly.
* Locking far away from the object is consider an anti-pattern that makes it 
hard to reason about locking by looking at the source only.

bq. do you think we should backport YARN-1043? Looks like an incompatible 
change so not sure whether we want it back to 1.0 line.

Always update all should be a compatible change, semantic wise, besides extra 
objects for non-changing metrics.  It seems that simply backporting the one 
line change (all is true always) from YARN-1043 (and keeping the test) is less 
risky to change the locking mechanisms.

> 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-10090.branch-1.2.patch, 
> HADOOP-10090.branch-1.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)

Reply via email to