Okay. Amar
On 6/2/10 2:31 PM, "Hong Tang (JIRA)" <[email protected]> wrote: [ https://issues.apache.org/jira/browse/MAPREDUCE-1833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12874512#action_12874512 ] Hong Tang commented on MAPREDUCE-1833: -------------------------------------- Hi Amar, Thanks for the comments. I think both suggestions are fair. Would you mind take over the jira? I did it as an emergency fix to get the 20.104 benchmarking done last week. I have created a ticket http://bug.corp.yahoo.com/show_bug.cgi?id=3703026. -Hong ------ Forwarded Message From: "Amar Kamat (JIRA)" <[email protected]> Date: Wed, 2 Jun 2010 04:43:38 -0400 To: Hong Tang <[email protected]> Subject: [jira] Commented: (MAPREDUCE-1833) [gridmix3] limit the maximum task duration in sleep job. [ https://issues.apache.org/jira/browse/MAPREDUCE-1833?page=com.atlassian.jira .plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12874502#act ion_12874502 ] Amar Kamat commented on MAPREDUCE-1833: --------------------------------------- Two comments 1) Shouldn't we keep GridMix.java (i.e GridMix command line params) independent of the synthetic job's (in this case SleepJob) config parameters? 2) Should we name the config parameter as gridmix.job.sleepjob.reduce.max-time instead of gridmix.job.sleepjob.reduce.max-time? This will allow us to put common job level configs under gridmix.job.* and a specific synthetic job level config under gridmix.job.jobname.* finish. Replaying such trace in sleep-job mode in Gridmix3 would unnecessarily prolong the benchmark execution time. It would be desirable to allow users to limit the maximum task duration. -- This message is automatically generated by JIRA. - You can reply to this email to add a comment to the issue online. ------ End of Forwarded Message > [gridmix3] limit the maximum task duration in sleep job. > -------------------------------------------------------- > > Key: MAPREDUCE-1833 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-1833 > Project: Hadoop Map/Reduce > Issue Type: Improvement > Components: contrib/gridmix > Reporter: Hong Tang > Attachments: mr-1833-yahoo-20.10.patch, mr-1833-yahoo-20.1xx.patch > > > In production job history logs, sometimes a task takes very long time to > finish. Replaying such trace in sleep-job mode in Gridmix3 would > unnecessarily prolong the benchmark execution time. It would be desirable to > allow users to limit the maximum task duration. -- This message is automatically generated by JIRA. - You can reply to this email to add a comment to the issue online.
