[ https://issues.apache.org/jira/browse/MAPREDUCE-5844?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14000413#comment-14000413 ]
Maysam Yabandeh commented on MAPREDUCE-5844: -------------------------------------------- [~jlowe], we observed more of these cases where the queue was actually full and a fix for proper headroom calculation would not help either. The thing is that although the queue might be full at each point of time, there is a constant flow of containers completing in the queue and new containers being assigned. Therefore, if the MRAppMaster does not aggressively preempt its reducer and ask for a container for its failed mapper, it is actually quite likely to get the mapper in a timely manner. Was chatting offline with [~kkambatl] and it came up that perhaps delayed preemption could be a more reasonable reaction in such cases. I was wondering what is your take on that? > Reducer Preemption is too aggressive > ------------------------------------ > > Key: MAPREDUCE-5844 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-5844 > Project: Hadoop Map/Reduce > Issue Type: Bug > Reporter: Maysam Yabandeh > Assignee: Maysam Yabandeh > > We observed cases where the reducer preemption makes the job finish much > later, and the preemption does not seem to be necessary since after > preemption both the preempted reducer and the mapper are assigned > immediately--meaning that there was already enough space for the mapper. > The logic for triggering preemption is at > RMContainerAllocator::preemptReducesIfNeeded > The preemption is triggered if the following is true: > {code} > headroom + am * |m| + pr * |r| < mapResourceRequest > {code} > where am: number of assigned mappers, |m| is mapper size, pr is number of > reducers being preempted, and |r| is the reducer size. > The original idea apparently was that if headroom is not big enough for the > new mapper requests, reducers should be preempted. This would work if the job > is alone in the cluster. Once we have queues, the headroom calculation > becomes more complicated and it would require a separate headroom calculation > per queue/job. > So, as a result headroom variable is kind of given up currently: *headroom is > always set to 0* What this implies to the speculation is that speculation > becomes very aggressive, not considering whether there is enough space for > the mappers or not. -- This message was sent by Atlassian JIRA (v6.2#6252)