[jira] [Created] (MAPREDUCE-7187) RMContainerAllocator.ScheduledRequests#getContainerReqToReplace may not find a task when the priority of container is PRIORITY_MAP
Zhizhen Hou created MAPREDUCE-7187: -- Summary: RMContainerAllocator.ScheduledRequests#getContainerReqToReplace may not find a task when the priority of container is PRIORITY_MAP Key: MAPREDUCE-7187 URL: https://issues.apache.org/jira/browse/MAPREDUCE-7187 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.7.5 Reporter: Zhizhen Hou The resource manager may has allocated a map container on a host ("h1" for example) for a application, and the container has not been fetched by the MRAppMaster. At this time, the MRAppMaster receives a task fail event, and the task is on host h1. The event cause the h1 blacklisted. Now the MRAppMaster send a heartbeat, and receive a container on h1. The MRAppMaster can not assign the container since it is on a blacklisted host. The #getContainerReqToReplace fails returning another task, may cause a map task hang forever. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: mapreduce-dev-unsubscr...@hadoop.apache.org For additional commands, e-mail: mapreduce-dev-h...@hadoop.apache.org
[jira] [Resolved] (MAPREDUCE-7081) Default speculator won't speculate the last several submitted reduced task if the total task num is large
[ https://issues.apache.org/jira/browse/MAPREDUCE-7081?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Zhizhen Hou resolved MAPREDUCE-7081. Resolution: Invalid > Default speculator won't speculate the last several submitted reduced task if > the total task num is large > - > > Key: MAPREDUCE-7081 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-7081 > Project: Hadoop Map/Reduce > Issue Type: Improvement > Components: mrv2 >Affects Versions: 2.9.0, 2.7.5 >Reporter: Zhizhen Hou >Priority: Major > > DefaultSpeculator speculates a task one time. By default, the number of > speculators is max(max(10, 0.01 * tasks.size), 0.1 * running tasks). > I set mapreduce.job.reduce.slowstart.completedmaps = 1 to start reduce after > all the map tasks are finished. The cluster has 1000 vcores, and the Job has > 5000 reduce jobs. At first, 1000 reduces tasks can run simultaneously, number > of speculators can speculator at most is 0.1 * 1000 = 100 tasks. Reduce tasks > with less data can over shortly, and speculator will speculator a task per > second by default. The task be speculated execution may be because the more > data to be processed. It will speculator 100 tasks within 100 seconds. When > 4900 reduces is over, If a reduce is executed with a lot of data be > processed and is put on a slow machine. The speculate opportunity is running > out, it will not be speculated. It can increase the execution time of job > significantly. > In short, it may waste the speculate opportunity at first only because the > execution time of reduce with less data to be processed as average time. At > end of job, there is no speculate opportunity available, especially last > several running tasks, judged the number of the running tasks . > In my opinion, the number of running tasks should not determine the number of > speculate opportunity .The number of tasks be speculated can be judged by > square of finished task percent. Take an example, if ninety percent of the > task is finished, only 0.9*0.9 = 0.81 speculate opportunity can be used. It > will leave enough opportunity for latter tasks. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: mapreduce-dev-unsubscr...@hadoop.apache.org For additional commands, e-mail: mapreduce-dev-h...@hadoop.apache.org
[jira] [Created] (MAPREDUCE-7081) Default speculator won't sepculate the last several submitted reduced task if the total task num is large
Zhizhen Hou created MAPREDUCE-7081: -- Summary: Default speculator won't sepculate the last several submitted reduced task if the total task num is large Key: MAPREDUCE-7081 URL: https://issues.apache.org/jira/browse/MAPREDUCE-7081 Project: Hadoop Map/Reduce Issue Type: Improvement Components: mrv2 Affects Versions: 2.7.5, 2.9.0 Reporter: Zhizhen Hou DefaultSpeculator speculates a task one time. By default, the number of speculators is max(max(10, 0.01 * tasks.size), 0.1 * running tasks). I set mapreduce.job.reduce.slowstart.completedmaps = 1 to start reduce after all the map tasks are finished. The cluster has 1000 vcores, and the Job has 5000 reduce jobs. At first, 1000 reduces tasks can run simultaneously, number of speculators can speculator at most is 0.1 * 1000 = 100 tasks. Reduce tasks with less data can over shortly, and speculator will speculator a task per second by default. The task be speculated execution may be because the more data to be processed. It will speculator 100 tasks within 100 seconds. When 4900 reduces is over, If a reduce is executed with a lot of data be processed and is put on a slow machine. The speculate opportunity is running out, it will not be speculated. It can increase the execution time of job significantly. In short, it may waste the speculate opportunity at first only because the execution time of reduce with less data to be processed as average time. At end of job, there is no speculate opportunity available, especially last several running tasks, judged the number of the running tasks . In my opinion, the number of running tasks should not determine the number of speculate opportunity .The number of tasks be speculated can be judged by square of finished task percent. Take an example, if ninety percent of the task is finished, only 0.9*0.9 = 0.81 speculate opportunity can be used. It will leave enough opportunity for latter tasks. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: mapreduce-dev-unsubscr...@hadoop.apache.org For additional commands, e-mail: mapreduce-dev-h...@hadoop.apache.org
[jira] [Created] (MAPREDUCE-7080) Default speculator won't sepculate the last several submitted reduced task if the total task num is large
Zhizhen Hou created MAPREDUCE-7080: -- Summary: Default speculator won't sepculate the last several submitted reduced task if the total task num is large Key: MAPREDUCE-7080 URL: https://issues.apache.org/jira/browse/MAPREDUCE-7080 Project: Hadoop Map/Reduce Issue Type: Improvement Components: mrv2 Affects Versions: 2.7.5 Reporter: Zhizhen Hou DefaultSpeculator speculates a task one time. By default, the number of speculators is max(max(10, 0.01 * tasks.size), 0.1 * running tasks) I set mapreduce.job.reduce.slowstart.completedmaps = 1 to start reduce after all the map tasks are finished. The cluster has 1000 vcores, and the Job has 5000 reduce jobs. At first, 1000 reduces tasks can run simultaneously, number of speculators can speculator at most is 0.1 * 1000 = 100 tasks. Reduce tasks with less data can over shortly, and speculator will speculator a task per second by default. The task be speculated execution may be because the more data to be processed. It will speculator 100 tasks within 100 seconds. When 4900 reduces is over, If a reduce is executed with a lot of data be processed and is put on a slow machine. The speculate opportunity is running out, it will not be speculated. It can increase the execution time of job significantly. In short, it may waste the speculate opportunity at first only because the execution time of reduce with less data to be processed as average time. At end of job, there is no speculate opportunity available, especially last several running tasks, judged the number of the running tasks . In my opinion, the number of tasks be speculated can be judged by square of finished task percent. Take an example, if ninety percent of the task is finished, only 0.9*0.9 = 0.81 speculate opportunity can be used. It will leave enough opportunity for latter tasks. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: mapreduce-dev-unsubscr...@hadoop.apache.org For additional commands, e-mail: mapreduce-dev-h...@hadoop.apache.org
[jira] [Created] (MAPREDUCE-7075) Change default configuration for mapreduce.reduce.shuffle.input.buffer.percent
Zhizhen Hou created MAPREDUCE-7075: -- Summary: Change default configuration for mapreduce.reduce.shuffle.input.buffer.percent Key: MAPREDUCE-7075 URL: https://issues.apache.org/jira/browse/MAPREDUCE-7075 Project: Hadoop Map/Reduce Issue Type: Improvement Components: mrv2 Affects Versions: 2.9.1 Reporter: Zhizhen Hou When the default value of mapreduce.reduce.shuffle.input.buffer.percent is 0.7, it may report OOM exception at shuffle stage. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: mapreduce-dev-unsubscr...@hadoop.apache.org For additional commands, e-mail: mapreduce-dev-h...@hadoop.apache.org