[jira] [Comment Edited] (MAPREDUCE-7227) Fix job staging directory residual problem in a big yarn cluster composed of multiple independent hdfs clusters
[ https://issues.apache.org/jira/browse/MAPREDUCE-7227?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17271674#comment-17271674 ] Eric Payne edited comment on MAPREDUCE-7227 at 1/25/21, 9:00 PM: - [~luoyuan], I'm sorry for the long delay. bq. I Set up two hdfs cluster, one named 'test-hdfs',another one named 'alg-hdfs', the test-hdfs also runs on yarn. So, IIUC, there is one YARN instance, but 2 HDFS instances, and YARN can use either one? And then each nodemanager would be configured to talk to both HDFS namenodes? I'm not certain YARN can support that. Each application in YARN can talk to another HDFS instance by specifying the full scheme to the namenode, -but I have not heard of the use case as you have described.- Is this setup a typical HDFS HA configuration? was (Author: eepayne): [~luoyuan], I'm sorry for the long delay. bq. I Set up two hdfs cluster, one named 'test-hdfs',another one named 'alg-hdfs', the test-hdfs also runs on yarn. So, IIUC, there is one YARN instance, but 2 HDFS instances, and YARN can use either one? And then each nodemanager would be configured to talk to both HDFS namenodes? I'm not certain YARN can support that. Each application in YARN can talk to another HDFS instance by specifying the full scheme to the namenode, but I have not heard of the use case as you have described. > Fix job staging directory residual problem in a big yarn cluster composed of > multiple independent hdfs clusters > --- > > Key: MAPREDUCE-7227 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-7227 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: applicationmaster, mrv2 >Affects Versions: 2.6.0, 2.7.0, 3.1.2 >Reporter: Yuan LUO >Assignee: Yuan LUO >Priority: Major > Attachments: 1.png, 2.png, HADOOP-MAPREDUCE-7227.001.patch, > HADOOP-MAPREDUCE-7227.002.patch, HADOOP-MAPREDUCE-7227.003.patch, > HADOOP-MAPREDUCE-7227.004.patch, HADOOP-MAPREDUCE-7227.005.patch, > Process_Analysis.png > > > Our yarn cluster is made up of some independent hdfs cluster, the > 'default.FS' in every hdfs cluster is different.when user submit job to yarn > cluster, if the 'default.FS' between client and nodemanager is > inconsistent, then the job staging dir can't be cleanup by AppMaster. Because > it will produce two job staging dirs in our conditions by client and > appmaster. So we can modify AppMaster through client's ‘default.FS’ to > create job staging dir. -- This message was sent by Atlassian Jira (v8.3.4#803005) - To unsubscribe, e-mail: mapreduce-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: mapreduce-issues-h...@hadoop.apache.org
[jira] [Commented] (MAPREDUCE-7227) Fix job staging directory residual problem in a big yarn cluster composed of multiple independent hdfs clusters
[ https://issues.apache.org/jira/browse/MAPREDUCE-7227?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17271674#comment-17271674 ] Eric Payne commented on MAPREDUCE-7227: --- [~luoyuan], I'm sorry for the long delay. bq. I Set up two hdfs cluster, one named 'test-hdfs',another one named 'alg-hdfs', the test-hdfs also runs on yarn. So, IIUC, there is one YARN instance, but 2 HDFS instances, and YARN can use either one? And then each nodemanager would be configured to talk to both HDFS namenodes? I'm not certain YARN can support that. Each application in YARN can talk to another HDFS instance by specifying the full scheme to the namenode, but I have not heard of the use case as you have described. > Fix job staging directory residual problem in a big yarn cluster composed of > multiple independent hdfs clusters > --- > > Key: MAPREDUCE-7227 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-7227 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: applicationmaster, mrv2 >Affects Versions: 2.6.0, 2.7.0, 3.1.2 >Reporter: Yuan LUO >Assignee: Yuan LUO >Priority: Major > Attachments: 1.png, 2.png, HADOOP-MAPREDUCE-7227.001.patch, > HADOOP-MAPREDUCE-7227.002.patch, HADOOP-MAPREDUCE-7227.003.patch, > HADOOP-MAPREDUCE-7227.004.patch, HADOOP-MAPREDUCE-7227.005.patch, > Process_Analysis.png > > > Our yarn cluster is made up of some independent hdfs cluster, the > 'default.FS' in every hdfs cluster is different.when user submit job to yarn > cluster, if the 'default.FS' between client and nodemanager is > inconsistent, then the job staging dir can't be cleanup by AppMaster. Because > it will produce two job staging dirs in our conditions by client and > appmaster. So we can modify AppMaster through client's ‘default.FS’ to > create job staging dir. -- This message was sent by Atlassian Jira (v8.3.4#803005) - To unsubscribe, e-mail: mapreduce-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: mapreduce-issues-h...@hadoop.apache.org
[jira] [Commented] (MAPREDUCE-7169) Speculative attempts should not run on the same node
[ https://issues.apache.org/jira/browse/MAPREDUCE-7169?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17271291#comment-17271291 ] Bilwa S T commented on MAPREDUCE-7169: -- Hi [~epayne] Can you please help in reviewing this patch? > Speculative attempts should not run on the same node > > > Key: MAPREDUCE-7169 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-7169 > Project: Hadoop Map/Reduce > Issue Type: New Feature > Components: yarn >Affects Versions: 2.7.2 >Reporter: Lee chen >Assignee: Bilwa S T >Priority: Major > Attachments: MAPREDUCE-7169-001.patch, MAPREDUCE-7169-002.patch, > MAPREDUCE-7169-003.patch, MAPREDUCE-7169.004.patch, MAPREDUCE-7169.005.patch, > MAPREDUCE-7169.006.patch, MAPREDUCE-7169.007.patch, > image-2018-12-03-09-54-07-859.png > > > I found in all versions of yarn, Speculative Execution may set the > speculative task to the node of original task.What i have read is only it > will try to have one more task attempt. haven't seen any place mentioning not > on same node.It is unreasonable.If the node have some problems lead to tasks > execution will be very slow. and then placement the speculative task to same > node cannot help the problematic task. > In our cluster （version 2.7.2，2700 nodes），this phenomenon appear > almost everyday. > !image-2018-12-03-09-54-07-859.png! -- This message was sent by Atlassian Jira (v8.3.4#803005) - To unsubscribe, e-mail: mapreduce-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: mapreduce-issues-h...@hadoop.apache.org
