[
https://issues.apache.org/jira/browse/MAPREDUCE-2137?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12922873#action_12922873
]
Ranjit Mathew commented on MAPREDUCE-2137:
------------------------------------------
Thanks for doing this. Some comments:
* We'll have to update the documentation accordingly, but perhaps only after
MAPREDUCE-1931 is committed.
* I prefer changing "JOBNAMEPREFIX" to "JOBNAME_PREFIX" or even "JOB_NAME_PFX"
so that it's more readable. Ditto for "ORIGJOBID" to "ORIG_JOBID" or
"ORIG_JOB_ID".
* The {{StringBuilder}} in {{initialValue()}} has an _initial capacity_ of 64,
but the comment makes it seem as if we're talking about the _total capacity_. I
suggest dropping that comment.
* I know that we're not making the GridMix job-name's format a contract, but do
you think it makes sense to check that the job has an expected format in the
unit-test? (Since Rumen does not generate traces containing the values
corresponding to "gridmix.job.id.original", the job name is the only link back
to the original job if you're looking at a Rumen-generated trace.)
> Mapping between Gridmix jobs and the corresponding original MR jobs is needed
> -----------------------------------------------------------------------------
>
> Key: MAPREDUCE-2137
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-2137
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: contrib/gridmix
> Affects Versions: 0.22.0
> Reporter: Ravi Gummadi
> Assignee: Ravi Gummadi
> Fix For: 0.22.0
>
> Attachments: 2137.patch
>
>
> Consider a trace file "trace1" obtained by running Rumen on a set of MR jobs'
> history logs. When gridmix runs simulated jobs from "trace1", it may skip
> some of the jobs from the trace file for some reason like out-of-order-jobs.
> Now use Rumen to generate trace2 from the history logs of gridmix's simulated
> jobs.
> Now, to compare and analyze the gridmix's simulated jobs with original MR
> jobs, we need a mapping between them.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
