[jira] Updated: (MAPREDUCE-1687) Stress submission policy does not always stress the cluster.
[ https://issues.apache.org/jira/browse/MAPREDUCE-1687?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Hong Tang updated MAPREDUCE-1687: - Attachment: mr-1687-yhadoop-20.1xx-20100423-2.patch Fixed a minor bug in StressReaderThread.run() where the condition of the while loop waiting for cluster to be underloaded is incorrect. > Stress submission policy does not always stress the cluster. > > > Key: MAPREDUCE-1687 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-1687 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: contrib/gridmix >Reporter: Hong Tang > Attachments: mr-1687-yhadoop-20.1xx-20100416.patch, > mr-1687-yhadoop-20.1xx-20100423-2.patch, mr-1687-yhadoop-20.1xx-20100423.patch > > > Currently, the rough idea of stress submission policy is to continue > submitting jobs until the pending map tasks reach 2x of the cluster capacity. > This proves to be inadequate and we saw a large job could monopolize the > whole cluster. -- This message is automatically generated by JIRA. - You can reply to this email to add a comment to the issue online.
[jira] Updated: (MAPREDUCE-1687) Stress submission policy does not always stress the cluster.
[ https://issues.apache.org/jira/browse/MAPREDUCE-1687?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Hong Tang updated MAPREDUCE-1687: - Attachment: mr-1687-yhadoop-20.1xx-20100423.patch Patch that fine tunes the overload determination heuristics. > Stress submission policy does not always stress the cluster. > > > Key: MAPREDUCE-1687 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-1687 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: contrib/gridmix >Reporter: Hong Tang > Attachments: mr-1687-yhadoop-20.1xx-20100416.patch, > mr-1687-yhadoop-20.1xx-20100423.patch > > > Currently, the rough idea of stress submission policy is to continue > submitting jobs until the pending map tasks reach 2x of the cluster capacity. > This proves to be inadequate and we saw a large job could monopolize the > whole cluster. -- This message is automatically generated by JIRA. - You can reply to this email to add a comment to the issue online.
[jira] Updated: (MAPREDUCE-1687) Stress submission policy does not always stress the cluster.
[ https://issues.apache.org/jira/browse/MAPREDUCE-1687?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Hong Tang updated MAPREDUCE-1687: - Attachment: mr-1687-yhadoop-20.1xx-20100416.patch Preliminary patch that caps the maximum number of incomplete map tasks for each job in cluster overload calculation. > Stress submission policy does not always stress the cluster. > > > Key: MAPREDUCE-1687 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-1687 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: contrib/gridmix >Reporter: Hong Tang > Attachments: mr-1687-yhadoop-20.1xx-20100416.patch > > > Currently, the rough idea of stress submission policy is to continue > submitting jobs until the pending map tasks reach 2x of the cluster capacity. > This proves to be inadequate and we saw a large job could monopolize the > whole cluster. -- This message is automatically generated by JIRA. - If you think it was sent incorrectly contact one of the administrators: https://issues.apache.org/jira/secure/Administrators.jspa - For more information on JIRA, see: http://www.atlassian.com/software/jira