[ https://issues.apache.org/jira/browse/MAPREDUCE-311?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14070802#comment-14070802 ]
Allen Wittenauer edited comment on MAPREDUCE-311 at 7/22/14 8:11 PM: --------------------------------------------------------------------- It's likely too late to change hadoop.tmp.dir. But this is still an issue. Debating opening a new JIRA (under YARN) that states the problem but not a solution so that hadoop.tmp.dir is left alone. was (Author: aw): It's likely too late to change hadoop.tmp.dir. But this is still an issue. Debating opening a new JIRA that states the problem but not a solution so that hadoop.tmp.dir is left alone. > JobClient should use multiple volumes as hadoop.tmp.dir > ------------------------------------------------------- > > Key: MAPREDUCE-311 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-311 > Project: Hadoop Map/Reduce > Issue Type: Improvement > Environment: All > Reporter: Milind Bhandarkar > > Currently, hadoop.tmp.dir configuration variable allows specification of only > a single directory to be used as scratch space. In particular, on the job > launcher nodes with multiple volumes, this fails the entire job if the > tmp.dir is somehow unusable. When the job launcher nodes have multiple > volumes, the tmp space availability can be improved by using multiple volumes > (either randomly or in round-robin.) The code for choosing a volume from a > comma-separated list of multiple volumes is already there for > mapred.local.dir etc. That needs to be used by job client as well. -- This message was sent by Atlassian JIRA (v6.2#6252)