[jira] [Commented] (MAPREDUCE-5817) Mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16506903#comment-16506903 ] Kartik Bhatia commented on MAPREDUCE-5817: -- [~jira.shegalov] Thanks. Will do that. > Mappers get rescheduled on node transition even after all reducers are > completed > > > Key: MAPREDUCE-5817 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: applicationmaster >Affects Versions: 2.3.0 >Reporter: Sangjin Lee >Assignee: Sangjin Lee >Priority: Major > Fix For: 2.8.0, 2.7.3, 2.6.5, 3.0.0-alpha1 > > Attachments: MAPREDUCE-5817.001.patch, MAPREDUCE-5817.002.patch, > mapreduce-5817.patch > > > We're seeing a behavior where a job runs long after all reducers were already > finished. We found that the job was rescheduling and running a number of > mappers beyond the point of reducer completion. In one situation, the job ran > for some 9 more hours after all reducers completed! > This happens because whenever a node transition (to an unusable state) comes > into the app master, it just reschedules all mappers that already ran on the > node in all cases. > Therefore, if any node transition has a potential to extend the job period. > Once this window opens, another node transition can prolong it, and this can > happen indefinitely in theory. > If there is some instability in the pool (unhealthy, etc.) for a duration, > then any big job is severely vulnerable to this problem. > If all reducers have been completed, JobImpl.actOnUnusableNode() should not > reschedule mapper tasks. If all reducers are completed, the mapper outputs > are no longer needed, and there is no need to reschedule mapper tasks as they > would not be consumed anyway. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: mapreduce-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: mapreduce-issues-h...@hadoop.apache.org
[jira] [Commented] (MAPREDUCE-5817) Mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16506427#comment-16506427 ] Gera Shegalov commented on MAPREDUCE-5817: -- [~kartik.bhatia], this Jira is done. You can file and work on the next iteration separately. It will require tracking shuffle completions in the task attempt state machine. > Mappers get rescheduled on node transition even after all reducers are > completed > > > Key: MAPREDUCE-5817 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: applicationmaster >Affects Versions: 2.3.0 >Reporter: Sangjin Lee >Assignee: Sangjin Lee >Priority: Major > Fix For: 2.8.0, 2.7.3, 2.6.5, 3.0.0-alpha1 > > Attachments: MAPREDUCE-5817.001.patch, MAPREDUCE-5817.002.patch, > mapreduce-5817.patch > > > We're seeing a behavior where a job runs long after all reducers were already > finished. We found that the job was rescheduling and running a number of > mappers beyond the point of reducer completion. In one situation, the job ran > for some 9 more hours after all reducers completed! > This happens because whenever a node transition (to an unusable state) comes > into the app master, it just reschedules all mappers that already ran on the > node in all cases. > Therefore, if any node transition has a potential to extend the job period. > Once this window opens, another node transition can prolong it, and this can > happen indefinitely in theory. > If there is some instability in the pool (unhealthy, etc.) for a duration, > then any big job is severely vulnerable to this problem. > If all reducers have been completed, JobImpl.actOnUnusableNode() should not > reschedule mapper tasks. If all reducers are completed, the mapper outputs > are no longer needed, and there is no need to reschedule mapper tasks as they > would not be consumed anyway. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: mapreduce-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: mapreduce-issues-h...@hadoop.apache.org
[jira] [Commented] (MAPREDUCE-5817) Mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16501426#comment-16501426 ] Kartik Bhatia commented on MAPREDUCE-5817: -- [~sjlee0] Instead of adding check allReducersComplete() in JobImpl can we not add check like allReducersRequiringMapComplete() that checks if all reducers have crossed shuffle phase we do not need to relaunch mappers on unusable node? > Mappers get rescheduled on node transition even after all reducers are > completed > > > Key: MAPREDUCE-5817 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: applicationmaster >Affects Versions: 2.3.0 >Reporter: Sangjin Lee >Assignee: Sangjin Lee >Priority: Major > Fix For: 2.8.0, 2.7.3, 2.6.5, 3.0.0-alpha1 > > Attachments: MAPREDUCE-5817.001.patch, MAPREDUCE-5817.002.patch, > mapreduce-5817.patch > > > We're seeing a behavior where a job runs long after all reducers were already > finished. We found that the job was rescheduling and running a number of > mappers beyond the point of reducer completion. In one situation, the job ran > for some 9 more hours after all reducers completed! > This happens because whenever a node transition (to an unusable state) comes > into the app master, it just reschedules all mappers that already ran on the > node in all cases. > Therefore, if any node transition has a potential to extend the job period. > Once this window opens, another node transition can prolong it, and this can > happen indefinitely in theory. > If there is some instability in the pool (unhealthy, etc.) for a duration, > then any big job is severely vulnerable to this problem. > If all reducers have been completed, JobImpl.actOnUnusableNode() should not > reschedule mapper tasks. If all reducers are completed, the mapper outputs > are no longer needed, and there is no need to reschedule mapper tasks as they > would not be consumed anyway. -- This message was sent by Atlassian JIRA (v7.6.3#76005) - To unsubscribe, e-mail: mapreduce-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: mapreduce-issues-h...@hadoop.apache.org
[jira] [Commented] (MAPREDUCE-5817) Mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15488750#comment-15488750 ] Chris Trezzo commented on MAPREDUCE-5817: - Thanks! > Mappers get rescheduled on node transition even after all reducers are > completed > > > Key: MAPREDUCE-5817 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: applicationmaster >Affects Versions: 2.3.0 >Reporter: Sangjin Lee >Assignee: Sangjin Lee > Fix For: 2.7.3, 2.6.5, 3.0.0-alpha1 > > Attachments: MAPREDUCE-5817.001.patch, MAPREDUCE-5817.002.patch, > mapreduce-5817.patch > > > We're seeing a behavior where a job runs long after all reducers were already > finished. We found that the job was rescheduling and running a number of > mappers beyond the point of reducer completion. In one situation, the job ran > for some 9 more hours after all reducers completed! > This happens because whenever a node transition (to an unusable state) comes > into the app master, it just reschedules all mappers that already ran on the > node in all cases. > Therefore, if any node transition has a potential to extend the job period. > Once this window opens, another node transition can prolong it, and this can > happen indefinitely in theory. > If there is some instability in the pool (unhealthy, etc.) for a duration, > then any big job is severely vulnerable to this problem. > If all reducers have been completed, JobImpl.actOnUnusableNode() should not > reschedule mapper tasks. If all reducers are completed, the mapper outputs > are no longer needed, and there is no need to reschedule mapper tasks as they > would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: mapreduce-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: mapreduce-issues-h...@hadoop.apache.org
[jira] [Commented] (MAPREDUCE-5817) Mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15431581#comment-15431581 ] Chris Trezzo commented on MAPREDUCE-5817: - Adding 2.6.5 to the target versions with the intention of backporting this to branch-2.6. Please let me know if you think otherwise. Thanks! > Mappers get rescheduled on node transition even after all reducers are > completed > > > Key: MAPREDUCE-5817 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: applicationmaster >Affects Versions: 2.3.0 >Reporter: Sangjin Lee >Assignee: Sangjin Lee > Fix For: 2.7.3 > > Attachments: MAPREDUCE-5817.001.patch, MAPREDUCE-5817.002.patch, > mapreduce-5817.patch > > > We're seeing a behavior where a job runs long after all reducers were already > finished. We found that the job was rescheduling and running a number of > mappers beyond the point of reducer completion. In one situation, the job ran > for some 9 more hours after all reducers completed! > This happens because whenever a node transition (to an unusable state) comes > into the app master, it just reschedules all mappers that already ran on the > node in all cases. > Therefore, if any node transition has a potential to extend the job period. > Once this window opens, another node transition can prolong it, and this can > happen indefinitely in theory. > If there is some instability in the pool (unhealthy, etc.) for a duration, > then any big job is severely vulnerable to this problem. > If all reducers have been completed, JobImpl.actOnUnusableNode() should not > reschedule mapper tasks. If all reducers are completed, the mapper outputs > are no longer needed, and there is no need to reschedule mapper tasks as they > would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332) - To unsubscribe, e-mail: mapreduce-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: mapreduce-issues-h...@hadoop.apache.org
[jira] [Commented] (MAPREDUCE-5817) Mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15248114#comment-15248114 ] Wangda Tan commented on MAPREDUCE-5817: --- Updated fix version. > Mappers get rescheduled on node transition even after all reducers are > completed > > > Key: MAPREDUCE-5817 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: applicationmaster >Affects Versions: 2.3.0 >Reporter: Sangjin Lee >Assignee: Sangjin Lee > Fix For: 2.7.3 > > Attachments: MAPREDUCE-5817.001.patch, MAPREDUCE-5817.002.patch, > mapreduce-5817.patch > > > We're seeing a behavior where a job runs long after all reducers were already > finished. We found that the job was rescheduling and running a number of > mappers beyond the point of reducer completion. In one situation, the job ran > for some 9 more hours after all reducers completed! > This happens because whenever a node transition (to an unusable state) comes > into the app master, it just reschedules all mappers that already ran on the > node in all cases. > Therefore, if any node transition has a potential to extend the job period. > Once this window opens, another node transition can prolong it, and this can > happen indefinitely in theory. > If there is some instability in the pool (unhealthy, etc.) for a duration, > then any big job is severely vulnerable to this problem. > If all reducers have been completed, JobImpl.actOnUnusableNode() should not > reschedule mapper tasks. If all reducers are completed, the mapper outputs > are no longer needed, and there is no need to reschedule mapper tasks as they > would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) Mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15247181#comment-15247181 ] Wangda Tan commented on MAPREDUCE-5817: --- Thanks [~sjlee0], committing now. > Mappers get rescheduled on node transition even after all reducers are > completed > > > Key: MAPREDUCE-5817 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: applicationmaster >Affects Versions: 2.3.0 >Reporter: Sangjin Lee >Assignee: Sangjin Lee > Fix For: 2.8.0 > > Attachments: MAPREDUCE-5817.001.patch, MAPREDUCE-5817.002.patch, > mapreduce-5817.patch > > > We're seeing a behavior where a job runs long after all reducers were already > finished. We found that the job was rescheduling and running a number of > mappers beyond the point of reducer completion. In one situation, the job ran > for some 9 more hours after all reducers completed! > This happens because whenever a node transition (to an unusable state) comes > into the app master, it just reschedules all mappers that already ran on the > node in all cases. > Therefore, if any node transition has a potential to extend the job period. > Once this window opens, another node transition can prolong it, and this can > happen indefinitely in theory. > If there is some instability in the pool (unhealthy, etc.) for a duration, > then any big job is severely vulnerable to this problem. > If all reducers have been completed, JobImpl.actOnUnusableNode() should not > reschedule mapper tasks. If all reducers are completed, the mapper outputs > are no longer needed, and there is no need to reschedule mapper tasks as they > would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) Mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15246018#comment-15246018 ] Sangjin Lee commented on MAPREDUCE-5817: I have no objections to backporting this to branch-2.7 or branch-2.6. Thanks! > Mappers get rescheduled on node transition even after all reducers are > completed > > > Key: MAPREDUCE-5817 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: applicationmaster >Affects Versions: 2.3.0 >Reporter: Sangjin Lee >Assignee: Sangjin Lee > Fix For: 2.8.0 > > Attachments: MAPREDUCE-5817.001.patch, MAPREDUCE-5817.002.patch, > mapreduce-5817.patch > > > We're seeing a behavior where a job runs long after all reducers were already > finished. We found that the job was rescheduling and running a number of > mappers beyond the point of reducer completion. In one situation, the job ran > for some 9 more hours after all reducers completed! > This happens because whenever a node transition (to an unusable state) comes > into the app master, it just reschedules all mappers that already ran on the > node in all cases. > Therefore, if any node transition has a potential to extend the job period. > Once this window opens, another node transition can prolong it, and this can > happen indefinitely in theory. > If there is some instability in the pool (unhealthy, etc.) for a duration, > then any big job is severely vulnerable to this problem. > If all reducers have been completed, JobImpl.actOnUnusableNode() should not > reschedule mapper tasks. If all reducers are completed, the mapper outputs > are no longer needed, and there is no need to reschedule mapper tasks as they > would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) Mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15242217#comment-15242217 ] Wangda Tan commented on MAPREDUCE-5817: --- Forgot to mention: MAPREDUCE-6513 depends on this patch. It can apply to branch-2.7 cleanly after backporting this patch. > Mappers get rescheduled on node transition even after all reducers are > completed > > > Key: MAPREDUCE-5817 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: applicationmaster >Affects Versions: 2.3.0 >Reporter: Sangjin Lee >Assignee: Sangjin Lee > Fix For: 2.8.0 > > Attachments: MAPREDUCE-5817.001.patch, MAPREDUCE-5817.002.patch, > mapreduce-5817.patch > > > We're seeing a behavior where a job runs long after all reducers were already > finished. We found that the job was rescheduling and running a number of > mappers beyond the point of reducer completion. In one situation, the job ran > for some 9 more hours after all reducers completed! > This happens because whenever a node transition (to an unusable state) comes > into the app master, it just reschedules all mappers that already ran on the > node in all cases. > Therefore, if any node transition has a potential to extend the job period. > Once this window opens, another node transition can prolong it, and this can > happen indefinitely in theory. > If there is some instability in the pool (unhealthy, etc.) for a duration, > then any big job is severely vulnerable to this problem. > If all reducers have been completed, JobImpl.actOnUnusableNode() should not > reschedule mapper tasks. If all reducers are completed, the mapper outputs > are no longer needed, and there is no need to reschedule mapper tasks as they > would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) Mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15242211#comment-15242211 ] Wangda Tan commented on MAPREDUCE-5817: --- [~sjlee0]/[~kasha], Should we backport this patch to branch-2.7? Any concerns? Thanks, > Mappers get rescheduled on node transition even after all reducers are > completed > > > Key: MAPREDUCE-5817 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: applicationmaster >Affects Versions: 2.3.0 >Reporter: Sangjin Lee >Assignee: Sangjin Lee > Fix For: 2.8.0 > > Attachments: MAPREDUCE-5817.001.patch, MAPREDUCE-5817.002.patch, > mapreduce-5817.patch > > > We're seeing a behavior where a job runs long after all reducers were already > finished. We found that the job was rescheduling and running a number of > mappers beyond the point of reducer completion. In one situation, the job ran > for some 9 more hours after all reducers completed! > This happens because whenever a node transition (to an unusable state) comes > into the app master, it just reschedules all mappers that already ran on the > node in all cases. > Therefore, if any node transition has a potential to extend the job period. > Once this window opens, another node transition can prolong it, and this can > happen indefinitely in theory. > If there is some instability in the pool (unhealthy, etc.) for a duration, > then any big job is severely vulnerable to this problem. > If all reducers have been completed, JobImpl.actOnUnusableNode() should not > reschedule mapper tasks. If all reducers are completed, the mapper outputs > are no longer needed, and there is no need to reschedule mapper tasks as they > would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) Mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14698218#comment-14698218 ] Hudson commented on MAPREDUCE-5817: --- FAILURE: Integrated in Hadoop-Yarn-trunk-Java8 #288 (See [https://builds.apache.org/job/Hadoop-Yarn-trunk-Java8/288/]) MAPREDUCE-5817. Mappers get rescheduled on node transition even after all reducers are completed. (Sangjin Lee via kasha) (kasha: rev 27d24f96ab8d17e839a1ef0d7076efc78d28724a) * hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app/src/test/java/org/apache/hadoop/mapreduce/v2/app/job/impl/TestJobImpl.java * hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app/src/main/java/org/apache/hadoop/mapreduce/v2/app/job/impl/JobImpl.java * hadoop-mapreduce-project/CHANGES.txt Mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Fix For: 2.8.0 Attachments: MAPREDUCE-5817.001.patch, MAPREDUCE-5817.002.patch, mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) Mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14698228#comment-14698228 ] Hudson commented on MAPREDUCE-5817: --- FAILURE: Integrated in Hadoop-Yarn-trunk #1018 (See [https://builds.apache.org/job/Hadoop-Yarn-trunk/1018/]) MAPREDUCE-5817. Mappers get rescheduled on node transition even after all reducers are completed. (Sangjin Lee via kasha) (kasha: rev 27d24f96ab8d17e839a1ef0d7076efc78d28724a) * hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app/src/test/java/org/apache/hadoop/mapreduce/v2/app/job/impl/TestJobImpl.java * hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app/src/main/java/org/apache/hadoop/mapreduce/v2/app/job/impl/JobImpl.java * hadoop-mapreduce-project/CHANGES.txt Mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Fix For: 2.8.0 Attachments: MAPREDUCE-5817.001.patch, MAPREDUCE-5817.002.patch, mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) Mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14698276#comment-14698276 ] Hudson commented on MAPREDUCE-5817: --- FAILURE: Integrated in Hadoop-Mapreduce-trunk-Java8 #285 (See [https://builds.apache.org/job/Hadoop-Mapreduce-trunk-Java8/285/]) MAPREDUCE-5817. Mappers get rescheduled on node transition even after all reducers are completed. (Sangjin Lee via kasha) (kasha: rev 27d24f96ab8d17e839a1ef0d7076efc78d28724a) * hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app/src/test/java/org/apache/hadoop/mapreduce/v2/app/job/impl/TestJobImpl.java * hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app/src/main/java/org/apache/hadoop/mapreduce/v2/app/job/impl/JobImpl.java * hadoop-mapreduce-project/CHANGES.txt Mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Fix For: 2.8.0 Attachments: MAPREDUCE-5817.001.patch, MAPREDUCE-5817.002.patch, mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) Mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14698294#comment-14698294 ] Hudson commented on MAPREDUCE-5817: --- FAILURE: Integrated in Hadoop-Hdfs-trunk-Java8 #277 (See [https://builds.apache.org/job/Hadoop-Hdfs-trunk-Java8/277/]) MAPREDUCE-5817. Mappers get rescheduled on node transition even after all reducers are completed. (Sangjin Lee via kasha) (kasha: rev 27d24f96ab8d17e839a1ef0d7076efc78d28724a) * hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app/src/test/java/org/apache/hadoop/mapreduce/v2/app/job/impl/TestJobImpl.java * hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app/src/main/java/org/apache/hadoop/mapreduce/v2/app/job/impl/JobImpl.java * hadoop-mapreduce-project/CHANGES.txt Mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Fix For: 2.8.0 Attachments: MAPREDUCE-5817.001.patch, MAPREDUCE-5817.002.patch, mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) Mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14698271#comment-14698271 ] Hudson commented on MAPREDUCE-5817: --- FAILURE: Integrated in Hadoop-Mapreduce-trunk #2234 (See [https://builds.apache.org/job/Hadoop-Mapreduce-trunk/2234/]) MAPREDUCE-5817. Mappers get rescheduled on node transition even after all reducers are completed. (Sangjin Lee via kasha) (kasha: rev 27d24f96ab8d17e839a1ef0d7076efc78d28724a) * hadoop-mapreduce-project/CHANGES.txt * hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app/src/main/java/org/apache/hadoop/mapreduce/v2/app/job/impl/JobImpl.java * hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app/src/test/java/org/apache/hadoop/mapreduce/v2/app/job/impl/TestJobImpl.java Mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Fix For: 2.8.0 Attachments: MAPREDUCE-5817.001.patch, MAPREDUCE-5817.002.patch, mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) Mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14698283#comment-14698283 ] Hudson commented on MAPREDUCE-5817: --- FAILURE: Integrated in Hadoop-Hdfs-trunk #2215 (See [https://builds.apache.org/job/Hadoop-Hdfs-trunk/2215/]) MAPREDUCE-5817. Mappers get rescheduled on node transition even after all reducers are completed. (Sangjin Lee via kasha) (kasha: rev 27d24f96ab8d17e839a1ef0d7076efc78d28724a) * hadoop-mapreduce-project/CHANGES.txt * hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app/src/main/java/org/apache/hadoop/mapreduce/v2/app/job/impl/JobImpl.java * hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app/src/test/java/org/apache/hadoop/mapreduce/v2/app/job/impl/TestJobImpl.java Mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Fix For: 2.8.0 Attachments: MAPREDUCE-5817.001.patch, MAPREDUCE-5817.002.patch, mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14697593#comment-14697593 ] Karthik Kambatla commented on MAPREDUCE-5817: - TestJobEndNotifier passes locally. +1, checking this in. mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Attachments: MAPREDUCE-5817.001.patch, MAPREDUCE-5817.002.patch, mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14697540#comment-14697540 ] Hadoop QA commented on MAPREDUCE-5817: -- \\ \\ | (x) *{color:red}-1 overall{color}* | \\ \\ || Vote || Subsystem || Runtime || Comment || | {color:blue}0{color} | pre-patch | 16m 2s | Pre-patch trunk compilation is healthy. | | {color:green}+1{color} | @author | 0m 0s | The patch does not contain any @author tags. | | {color:green}+1{color} | tests included | 0m 0s | The patch appears to include 1 new or modified test files. | | {color:green}+1{color} | javac | 7m 40s | There were no new javac warning messages. | | {color:green}+1{color} | javadoc | 9m 51s | There were no new javadoc warning messages. | | {color:green}+1{color} | release audit | 0m 23s | The applied patch does not increase the total number of release audit warnings. | | {color:red}-1{color} | checkstyle | 0m 33s | The applied patch generated 1 new checkstyle issues (total was 108, now 107). | | {color:green}+1{color} | whitespace | 0m 0s | The patch has no lines that end in whitespace. | | {color:green}+1{color} | install | 1m 23s | mvn install still works. | | {color:green}+1{color} | eclipse:eclipse | 0m 33s | The patch built with eclipse:eclipse. | | {color:green}+1{color} | findbugs | 1m 7s | The patch does not introduce any new Findbugs (version 3.0.0) warnings. | | {color:red}-1{color} | mapreduce tests | 8m 56s | Tests failed in hadoop-mapreduce-client-app. | | | | 46m 33s | | \\ \\ || Reason || Tests || | Failed unit tests | hadoop.mapreduce.v2.app.TestJobEndNotifier | \\ \\ || Subsystem || Report/Notes || | Patch URL | http://issues.apache.org/jira/secure/attachment/12750568/MAPREDUCE-5817.002.patch | | Optional Tests | javadoc javac unit findbugs checkstyle | | git revision | trunk / 84bf712 | | checkstyle | https://builds.apache.org/job/PreCommit-MAPREDUCE-Build/5936/artifact/patchprocess/diffcheckstylehadoop-mapreduce-client-app.txt | | hadoop-mapreduce-client-app test log | https://builds.apache.org/job/PreCommit-MAPREDUCE-Build/5936/artifact/patchprocess/testrun_hadoop-mapreduce-client-app.txt | | Test Results | https://builds.apache.org/job/PreCommit-MAPREDUCE-Build/5936/testReport/ | | Java | 1.7.0_55 | | uname | Linux asf903.gq1.ygridcore.net 3.13.0-36-lowlatency #63-Ubuntu SMP PREEMPT Wed Sep 3 21:56:12 UTC 2014 x86_64 x86_64 x86_64 GNU/Linux | | Console output | https://builds.apache.org/job/PreCommit-MAPREDUCE-Build/5936/console | This message was automatically generated. mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Attachments: MAPREDUCE-5817.001.patch, MAPREDUCE-5817.002.patch, mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) Mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14697606#comment-14697606 ] Hudson commented on MAPREDUCE-5817: --- FAILURE: Integrated in Hadoop-trunk-Commit #8306 (See [https://builds.apache.org/job/Hadoop-trunk-Commit/8306/]) MAPREDUCE-5817. Mappers get rescheduled on node transition even after all reducers are completed. (Sangjin Lee via kasha) (kasha: rev 27d24f96ab8d17e839a1ef0d7076efc78d28724a) * hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app/src/main/java/org/apache/hadoop/mapreduce/v2/app/job/impl/JobImpl.java * hadoop-mapreduce-project/CHANGES.txt * hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app/src/test/java/org/apache/hadoop/mapreduce/v2/app/job/impl/TestJobImpl.java Mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Attachments: MAPREDUCE-5817.001.patch, MAPREDUCE-5817.002.patch, mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) Mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14698112#comment-14698112 ] Sangjin Lee commented on MAPREDUCE-5817: Thanks [~chris.douglas] for the review and [~kasha] for the commit! Mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Fix For: 2.8.0 Attachments: MAPREDUCE-5817.001.patch, MAPREDUCE-5817.002.patch, mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14697284#comment-14697284 ] Ben Podgursky commented on MAPREDUCE-5817: -- Thanks Sangjin, that looks good to me (not that I'm super familiar with the hadoop source). mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Attachments: MAPREDUCE-5817.001.patch, mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14693807#comment-14693807 ] Chris Douglas commented on MAPREDUCE-5817: -- bq. The current patch skips re-running mappers only if all reducers are complete. So I don't think reducers will fail beyond that point? Did I understand your question right? I see; sorry, I hadn't read the rest of the JIRA carefully. That's a fairly narrow window, isn't it? We may not need an extra state, if we kill all running maps when the last reducer completes. The condition this adds prevents new maps from being scheduled while cleanup/commit code is running. Minor: could {{allReducersComplete()}} call {{getCompletedReduces()}}? +1 on the patch mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Attachments: MAPREDUCE-5817.001.patch, mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14692298#comment-14692298 ] Chris Douglas commented on MAPREDUCE-5817: -- Does this work if the reducer fails subsequently? Presumably reexecution will be triggered by fetch failures? mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Labels: BB2015-05-TBR Attachments: MAPREDUCE-5817.001.patch, mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14692599#comment-14692599 ] Sangjin Lee commented on MAPREDUCE-5817: The test failures appear unrelated. The checkstyle is about the length of file {{JobImpl.java}} which is pretty much an existing issue. mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Attachments: MAPREDUCE-5817.001.patch, mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14692507#comment-14692507 ] Hadoop QA commented on MAPREDUCE-5817: -- \\ \\ | (x) *{color:red}-1 overall{color}* | \\ \\ || Vote || Subsystem || Runtime || Comment || | {color:blue}0{color} | pre-patch | 16m 50s | Pre-patch trunk compilation is healthy. | | {color:green}+1{color} | @author | 0m 0s | The patch does not contain any @author tags. | | {color:green}+1{color} | tests included | 0m 0s | The patch appears to include 1 new or modified test files. | | {color:green}+1{color} | javac | 8m 11s | There were no new javac warning messages. | | {color:green}+1{color} | javadoc | 10m 4s | There were no new javadoc warning messages. | | {color:green}+1{color} | release audit | 0m 24s | The applied patch does not increase the total number of release audit warnings. | | {color:red}-1{color} | checkstyle | 0m 35s | The applied patch generated 1 new checkstyle issues (total was 108, now 107). | | {color:green}+1{color} | whitespace | 0m 0s | The patch has no lines that end in whitespace. | | {color:green}+1{color} | install | 1m 26s | mvn install still works. | | {color:green}+1{color} | eclipse:eclipse | 0m 35s | The patch built with eclipse:eclipse. | | {color:green}+1{color} | findbugs | 1m 10s | The patch does not introduce any new Findbugs (version 3.0.0) warnings. | | {color:red}-1{color} | mapreduce tests | 9m 6s | Tests failed in hadoop-mapreduce-client-app. | | | | 48m 26s | | \\ \\ || Reason || Tests || | Failed unit tests | hadoop.mapreduce.v2.app.TestJobEndNotifier | \\ \\ || Subsystem || Report/Notes || | Patch URL | http://issues.apache.org/jira/secure/attachment/12749938/MAPREDUCE-5817.001.patch | | Optional Tests | javadoc javac unit findbugs checkstyle | | git revision | trunk / 7c796fd | | checkstyle | https://builds.apache.org/job/PreCommit-MAPREDUCE-Build/5934/artifact/patchprocess/diffcheckstylehadoop-mapreduce-client-app.txt | | hadoop-mapreduce-client-app test log | https://builds.apache.org/job/PreCommit-MAPREDUCE-Build/5934/artifact/patchprocess/testrun_hadoop-mapreduce-client-app.txt | | Test Results | https://builds.apache.org/job/PreCommit-MAPREDUCE-Build/5934/testReport/ | | Java | 1.7.0_55 | | uname | Linux asf905.gq1.ygridcore.net 3.13.0-36-lowlatency #63-Ubuntu SMP PREEMPT Wed Sep 3 21:56:12 UTC 2014 x86_64 x86_64 x86_64 GNU/Linux | | Console output | https://builds.apache.org/job/PreCommit-MAPREDUCE-Build/5934/console | This message was automatically generated. mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Attachments: MAPREDUCE-5817.001.patch, mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14692455#comment-14692455 ] Sangjin Lee commented on MAPREDUCE-5817: The current patch skips re-running mappers only if all reducers are complete. So I don't think reducers will fail beyond that point? Did I understand your question right? mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Labels: BB2015-05-TBR Attachments: MAPREDUCE-5817.001.patch, mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14680580#comment-14680580 ] Ben Podgursky commented on MAPREDUCE-5817: -- Is this ticket on any roadmaps? We're running into a problem which I think is an extreme of this case -- we have a lot of MR jobs which are Map-only, and when a NodeManager goes unhealthy while other map tasks are finishing, the Map task is re-run, even though there are no reducers at all. If the tasks are slow, this is a huge waste of time. mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Labels: BB2015-05-TBR Attachments: mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14680903#comment-14680903 ] Sangjin Lee commented on MAPREDUCE-5817: Thanks for the comments [~bpodgursky]. I've been meaning to revisit this JIRA. This JIRA is stalled because there are 2 options to move forward with different implications. Option (1) is to prevent re-running mappers as soon as all reducers complete, and the option (2) is to add a new task attempt state to be able to handle a situation where a node becomes unstable while reducers are past the copy phase. As mentioned above, option (1) is narrower and simpler in scope, but it will ensure that the job will complete in time only after all reducers are complete. Option (2) can narrow this window further so that as soon as the we're past the copy phase we can reasonably ensure that the job will complete in time, but at the expense of a fair amount of complexity of the changes. FYI, our company has been running with option (1) for a while, and we've been more or less happy with that. I'd like to hear others' thoughts on this. The patch can be updated for the option (1) fairly quickly. It will take more time to complete option (2) on the other hand. We can discuss which option we want to go with, and even going with option (1) first in this JIRA and do option (2) in another JIRA. Thoughts? mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Labels: BB2015-05-TBR Attachments: mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14680708#comment-14680708 ] Hadoop QA commented on MAPREDUCE-5817: -- \\ \\ | (x) *{color:red}-1 overall{color}* | \\ \\ || Vote || Subsystem || Runtime || Comment || | {color:blue}0{color} | pre-patch | 15m 52s | Pre-patch trunk compilation is healthy. | | {color:green}+1{color} | @author | 0m 0s | The patch does not contain any @author tags. | | {color:green}+1{color} | tests included | 0m 0s | The patch appears to include 1 new or modified test files. | | {color:green}+1{color} | javac | 7m 41s | There were no new javac warning messages. | | {color:green}+1{color} | javadoc | 9m 40s | There were no new javadoc warning messages. | | {color:green}+1{color} | release audit | 0m 21s | The applied patch does not increase the total number of release audit warnings. | | {color:red}-1{color} | checkstyle | 0m 34s | The applied patch generated 3 new checkstyle issues (total was 108, now 109). | | {color:green}+1{color} | whitespace | 0m 0s | The patch has no lines that end in whitespace. | | {color:green}+1{color} | install | 1m 21s | mvn install still works. | | {color:green}+1{color} | eclipse:eclipse | 0m 31s | The patch built with eclipse:eclipse. | | {color:green}+1{color} | findbugs | 1m 4s | The patch does not introduce any new Findbugs (version 3.0.0) warnings. | | {color:red}-1{color} | mapreduce tests | 8m 52s | Tests failed in hadoop-mapreduce-client-app. | | | | 46m 0s | | \\ \\ || Reason || Tests || | Failed unit tests | hadoop.mapreduce.v2.app.TestJobEndNotifier | \\ \\ || Subsystem || Report/Notes || | Patch URL | http://issues.apache.org/jira/secure/attachment/12638107/mapreduce-5817.patch | | Optional Tests | javadoc javac unit findbugs checkstyle | | git revision | trunk / 8f73bdd | | checkstyle | https://builds.apache.org/job/PreCommit-MAPREDUCE-Build/5933/artifact/patchprocess/diffcheckstylehadoop-mapreduce-client-app.txt | | hadoop-mapreduce-client-app test log | https://builds.apache.org/job/PreCommit-MAPREDUCE-Build/5933/artifact/patchprocess/testrun_hadoop-mapreduce-client-app.txt | | Test Results | https://builds.apache.org/job/PreCommit-MAPREDUCE-Build/5933/testReport/ | | Java | 1.7.0_55 | | uname | Linux asf907.gq1.ygridcore.net 3.13.0-36-lowlatency #63-Ubuntu SMP PREEMPT Wed Sep 3 21:56:12 UTC 2014 x86_64 x86_64 x86_64 GNU/Linux | | Console output | https://builds.apache.org/job/PreCommit-MAPREDUCE-Build/5933/console | This message was automatically generated. mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Labels: BB2015-05-TBR Attachments: mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14681045#comment-14681045 ] Ben Podgursky commented on MAPREDUCE-5817: -- Option (2) sounds a lot trickier (and more of a risk tradeoff) because tasks do realistically die or get pre-empted after copying, and it could hurt runtime to * not * restart tasks in case that happens. So you'd want that to be configurable either way. In an ideal world, I would just kill all map tasks which were running at the time the last reduce finished... but I don't know the implications this would have for counters, etc. So I'd vote (1) for now since I think it's pure upside, and leave (2) for a later JIRA since involves more tradeoffs. mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Labels: BB2015-05-TBR Attachments: mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14524854#comment-14524854 ] Hadoop QA commented on MAPREDUCE-5817: -- \\ \\ | (x) *{color:red}-1 overall{color}* | \\ \\ || Vote || Subsystem || Runtime || Comment || | {color:red}-1{color} | patch | 0m 0s | The patch command could not apply the patch during dryrun. | \\ \\ || Subsystem || Report/Notes || | Patch URL | http://issues.apache.org/jira/secure/attachment/12638107/mapreduce-5817.patch | | Optional Tests | javadoc javac unit findbugs checkstyle | | git revision | trunk / f1a152c | | Console output | https://builds.apache.org/job/PreCommit-MAPREDUCE-Build/5557/console | This message was automatically generated. mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Attachments: mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=14524910#comment-14524910 ] Hadoop QA commented on MAPREDUCE-5817: -- \\ \\ | (x) *{color:red}-1 overall{color}* | \\ \\ || Vote || Subsystem || Runtime || Comment || | {color:red}-1{color} | patch | 0m 0s | The patch command could not apply the patch during dryrun. | \\ \\ || Subsystem || Report/Notes || | Patch URL | http://issues.apache.org/jira/secure/attachment/12638107/mapreduce-5817.patch | | Optional Tests | javadoc javac unit findbugs checkstyle | | git revision | trunk / f1a152c | | Console output | https://builds.apache.org/job/PreCommit-MAPREDUCE-Build/5566/console | This message was automatically generated. mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Attachments: mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.3.4#6332)
[jira] [Commented] (MAPREDUCE-5817) mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13957474#comment-13957474 ] Gera Shegalov commented on MAPREDUCE-5817: -- The scope is: We should redefine {{JobImpl.checkReadyForCommit}} to return {{COMMITTING}} when {code} if (numReduceTasks 0) { if (isucceededReduceTaskCount == numReduceTasks) return COMMITTING; } else if (completedTaskCount == tasks.size()) { return COMMITTING; } {code} To address unprecedented nature, we can introduce a new state for TaskAttempImpl LOST_COMPLETE that only mappers can go into from SUCCEEDED on lost node. On successful Job commit, we should generate an event that will kill all outstanding unfinished map task attempts. Move one map task attempt form LOST to SUCCEEDED (e.g, with minimum id) if there is none yet. mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Attachments: mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.2#6252)
[jira] [Commented] (MAPREDUCE-5817) mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13956786#comment-13956786 ] Sangjin Lee commented on MAPREDUCE-5817: I am leaning towards option (1) for its simplicity and smaller impact in general. It still leaves rescheduled mappers running when all reducers complete, but I think it would be a much smaller risk than the problem we're facing. I'll code up a patch and submit it soon. Let me know if you have suggestions/comments. mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.2#6252)
[jira] [Commented] (MAPREDUCE-5817) mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13956926#comment-13956926 ] Hadoop QA commented on MAPREDUCE-5817: -- {color:red}-1 overall{color}. Here are the results of testing the latest attachment http://issues.apache.org/jira/secure/attachment/12638107/mapreduce-5817.patch against trunk revision . {color:green}+1 @author{color}. The patch does not contain any @author tags. {color:green}+1 tests included{color}. The patch appears to include 1 new or modified test files. {color:green}+1 javac{color}. The applied patch does not increase the total number of javac compiler warnings. {color:green}+1 javadoc{color}. There were no new javadoc warning messages. {color:green}+1 eclipse:eclipse{color}. The patch built with eclipse:eclipse. {color:green}+1 findbugs{color}. The patch does not introduce any new Findbugs (version 1.3.9) warnings. {color:green}+1 release audit{color}. The applied patch does not increase the total number of release audit warnings. {color:red}-1 core tests{color}. The patch failed these unit tests in hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app: org.apache.hadoop.mapreduce.v2.app.TestMRAppMaster {color:green}+1 contrib tests{color}. The patch passed contrib unit tests. Test results: https://builds.apache.org/job/PreCommit-MAPREDUCE-Build/4476//testReport/ Console output: https://builds.apache.org/job/PreCommit-MAPREDUCE-Build/4476//console This message is automatically generated. mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Attachments: mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.2#6252)
[jira] [Commented] (MAPREDUCE-5817) mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13956998#comment-13956998 ] Sangjin Lee commented on MAPREDUCE-5817: The test failures are unrelated to this patch. They are coming from MAPREDUCE-5815. mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Attachments: mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.2#6252)
[jira] [Commented] (MAPREDUCE-5817) mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13957022#comment-13957022 ] Gera Shegalov commented on MAPREDUCE-5817: -- Thanks for working on this, [~sjlee0]. I would like to advocate for Option 2 with resurrect where the job moves to COMMITTING once all the output is in HDFS a) the job succeeds faster b) there is no ambiguity what mappers' output was actually consumed. mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Attachments: mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.2#6252)
[jira] [Commented] (MAPREDUCE-5817) mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13957127#comment-13957127 ] Sangjin Lee commented on MAPREDUCE-5817: [~jira.shegalov] I agree with the pros of option (2). On the other hand, I do feel uneasy about resurrecting a killed attempt which is necessary with option (2). I don't think that's done today, so it would be somewhat unprecedented. Also, what do you think the scope of changes would be? mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee Assignee: Sangjin Lee Attachments: mapreduce-5817.patch We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.2#6252)
[jira] [Commented] (MAPREDUCE-5817) mappers get rescheduled on node transition even after all reducers are completed
[ https://issues.apache.org/jira/browse/MAPREDUCE-5817?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanelfocusedCommentId=13955750#comment-13955750 ] Sangjin Lee commented on MAPREDUCE-5817: We're talking about two options for this: (1) modify JobImpl.actOnUnusableNode() so that if all reducers are completed do not reschedule mappers, and (2) modify checkReadyForCommit() so that it transitions to COMMITTING if all reducers are completed (if reducers exist) instead of checking all tasks are completed. Either approach seems to have some downsides. For (1), the change is pretty narrow (only affects the rescheduling scenario). However, it still lets the mapper tasks that were rescheduled prior to reducer completion run. So the job may linger until those mapper tasks run to completion. And if those mapper tasks fail for any reason, it may render the job as failed (even though all reducers may have succeeded in reality). For (2), it would be effective and would make the job finish much more quickly. On the other hand, we'd need to do something about the mapper tasks that are running at that point. They may need to be killed. Also, if the original mapper tasks were successful, we may need to resurrect their status from KILLED to SUCCESSFUL to avoid confusion. mappers get rescheduled on node transition even after all reducers are completed Key: MAPREDUCE-5817 URL: https://issues.apache.org/jira/browse/MAPREDUCE-5817 Project: Hadoop Map/Reduce Issue Type: Bug Components: applicationmaster Affects Versions: 2.3.0 Reporter: Sangjin Lee We're seeing a behavior where a job runs long after all reducers were already finished. We found that the job was rescheduling and running a number of mappers beyond the point of reducer completion. In one situation, the job ran for some 9 more hours after all reducers completed! This happens because whenever a node transition (to an unusable state) comes into the app master, it just reschedules all mappers that already ran on the node in all cases. Therefore, if any node transition has a potential to extend the job period. Once this window opens, another node transition can prolong it, and this can happen indefinitely in theory. If there is some instability in the pool (unhealthy, etc.) for a duration, then any big job is severely vulnerable to this problem. If all reducers have been completed, JobImpl.actOnUnusableNode() should not reschedule mapper tasks. If all reducers are completed, the mapper outputs are no longer needed, and there is no need to reschedule mapper tasks as they would not be consumed anyway. -- This message was sent by Atlassian JIRA (v6.2#6252)