[jira] [Commented] (MAPREDUCE-4659) Confusing output when running "hadoop version" from one hadoop installation when HADOOP_HOME points to another
[ https://issues.apache.org/jira/browse/MAPREDUCE-4659?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13456301#comment-13456301 ] Harsh J commented on MAPREDUCE-4659: Good idea, I've seen this bite a few users before. Note though, that on 2.x and current 1.x, the HADOOP_HOME was deprecated for HADOOP_PREFIX and per-component homes. So maybe we could also look at HADOOP_COMMON_HOME, which is the lib set that has the VersionInfo? > Confusing output when running "hadoop version" from one hadoop installation > when HADOOP_HOME points to another > -- > > Key: MAPREDUCE-4659 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-4659 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: client >Affects Versions: 0.20.2 >Reporter: Sandy Ryza > > Hadoop version X is downloaded to ~/hadoop-x, and Hadoop version Y is > downloaded to ~/hadoop-y. HADOOP_HOME is set to hadoop-x. A user running > hadoop-y/bin/hadoop might expect to be running the hadoop-y jars, but, > because of HADOOP_HOME, will actually be running hadoop-x jars. > "hadoop version" could help clear this up a little by reporting the current > HADOOP_HOME. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (MAPREDUCE-4647) We should only unjar jobjar if there is a lib directory in it.
[ https://issues.apache.org/jira/browse/MAPREDUCE-4647?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13456191#comment-13456191 ] Hadoop QA commented on MAPREDUCE-4647: -- +1 overall. Here are the results of testing the latest attachment http://issues.apache.org/jira/secure/attachment/12545216/MR-4647.txt against trunk revision . +1 @author. The patch does not contain any @author tags. +1 tests included. The patch appears to include 5 new or modified test files. +1 javac. The applied patch does not increase the total number of javac compiler warnings. +1 javadoc. The javadoc tool did not generate any warning messages. +1 eclipse:eclipse. The patch built with eclipse:eclipse. +1 findbugs. The patch does not introduce any new Findbugs (version 1.3.9) warnings. +1 release audit. The applied patch does not increase the total number of release audit warnings. +1 core tests. The patch passed unit tests in hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-app hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-common hadoop-mapreduce-project/hadoop-mapreduce-client/hadoop-mapreduce-client-jobclient hadoop-yarn-project/hadoop-yarn/hadoop-yarn-api hadoop-yarn-project/hadoop-yarn/hadoop-yarn-common hadoop-yarn-project/hadoop-yarn/hadoop-yarn-server/hadoop-yarn-server-nodemanager. +1 contrib tests. The patch passed contrib unit tests. Test results: https://builds.apache.org/job/PreCommit-MAPREDUCE-Build/2855//testReport/ Console output: https://builds.apache.org/job/PreCommit-MAPREDUCE-Build/2855//console This message is automatically generated. > We should only unjar jobjar if there is a lib directory in it. > -- > > Key: MAPREDUCE-4647 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-4647 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: mrv2 >Affects Versions: 0.23.3 >Reporter: Robert Joseph Evans >Assignee: Robert Joseph Evans > Attachments: MR-4647-branch-0.23.txt, MR-4647.txt, MR-4647.txt, > MR-4647.txt, MR-4647.txt > > > For backwards compatibility we recently added made is so we would unjar the > job.jar and add anything to the classpath in the lib directory of that jar. > But this also slows job startup down a lot if the jar is large. We should > only unjar it if actually doing so would add something new to the classpath. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (MAPREDUCE-4647) We should only unjar jobjar if there is a lib directory in it.
[ https://issues.apache.org/jira/browse/MAPREDUCE-4647?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Robert Joseph Evans updated MAPREDUCE-4647: --- Attachment: MR-4647.txt MR-4647-branch-0.23.txt Attaching an updated version for trunk/branch-2 and another for branch-0.23. These address the comments from before. > We should only unjar jobjar if there is a lib directory in it. > -- > > Key: MAPREDUCE-4647 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-4647 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: mrv2 >Affects Versions: 0.23.3 >Reporter: Robert Joseph Evans >Assignee: Robert Joseph Evans > Attachments: MR-4647-branch-0.23.txt, MR-4647.txt, MR-4647.txt, > MR-4647.txt, MR-4647.txt > > > For backwards compatibility we recently added made is so we would unjar the > job.jar and add anything to the classpath in the lib directory of that jar. > But this also slows job startup down a lot if the jar is large. We should > only unjar it if actually doing so would add something new to the classpath. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (MAPREDUCE-4647) We should only unjar jobjar if there is a lib directory in it.
[ https://issues.apache.org/jira/browse/MAPREDUCE-4647?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13456105#comment-13456105 ] Robert Joseph Evans commented on MAPREDUCE-4647: I'll look at updating the documentation, and the warning. Yes there is a minor difference between 0.23 and 2.0. The original patch applied to 0.23, but not 2.0 so I rebased on trunk. I will supply a 0.23 version with the updates as well. > We should only unjar jobjar if there is a lib directory in it. > -- > > Key: MAPREDUCE-4647 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-4647 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: mrv2 >Affects Versions: 0.23.3 >Reporter: Robert Joseph Evans >Assignee: Robert Joseph Evans > Attachments: MR-4647.txt, MR-4647.txt, MR-4647.txt > > > For backwards compatibility we recently added made is so we would unjar the > job.jar and add anything to the classpath in the lib directory of that jar. > But this also slows job startup down a lot if the jar is large. We should > only unjar it if actually doing so would add something new to the classpath. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (MAPREDUCE-4645) Providing a random seed to Slive should make the sequence of filenames completely deterministic
[ https://issues.apache.org/jira/browse/MAPREDUCE-4645?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Ravi Prakash updated MAPREDUCE-4645: Attachment: MAPREDUCE-4645.branch-0.23.patch Thanks for your review and suggestion Konstantin! I've updated the patch to use the taskID to seed the RNG. > Providing a random seed to Slive should make the sequence of filenames > completely deterministic > --- > > Key: MAPREDUCE-4645 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-4645 > Project: Hadoop Map/Reduce > Issue Type: Improvement > Components: performance, test >Affects Versions: 0.23.1, 2.0.0-alpha >Reporter: Ravi Prakash >Assignee: Ravi Prakash > Labels: performance, test > Attachments: MAPREDUCE-4645.branch-0.23.patch, > MAPREDUCE-4645.branch-0.23.patch > > > Using the -random seed option still doesn't produce a deterministic sequence > of filenames. Hence there's no way to replicate the performance test. If I'm > providing a seed, its obvious that I want the test to be reproducible. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (MAPREDUCE-4647) We should only unjar jobjar if there is a lib directory in it.
[ https://issues.apache.org/jira/browse/MAPREDUCE-4647?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13456065#comment-13456065 ] Thomas Graves commented on MAPREDUCE-4647: -- Few minor comments. - Since Pattern is really only implemented for jars but it still handles zip, tar.gz, tar we should try to document the behavior better. - It might be nice to add something like "even though specified as a Pattern" to the warning message in unpack in FSDownload.java > We should only unjar jobjar if there is a lib directory in it. > -- > > Key: MAPREDUCE-4647 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-4647 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: mrv2 >Affects Versions: 0.23.3 >Reporter: Robert Joseph Evans >Assignee: Robert Joseph Evans > Attachments: MR-4647.txt, MR-4647.txt, MR-4647.txt > > > For backwards compatibility we recently added made is so we would unjar the > job.jar and add anything to the classpath in the lib directory of that jar. > But this also slows job startup down a lot if the jar is large. We should > only unjar it if actually doing so would add something new to the classpath. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (MAPREDUCE-4647) We should only unjar jobjar if there is a lib directory in it.
[ https://issues.apache.org/jira/browse/MAPREDUCE-4647?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13456067#comment-13456067 ] Thomas Graves commented on MAPREDUCE-4647: -- Also your patch doesn't apply to branch-0.23 > We should only unjar jobjar if there is a lib directory in it. > -- > > Key: MAPREDUCE-4647 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-4647 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: mrv2 >Affects Versions: 0.23.3 >Reporter: Robert Joseph Evans >Assignee: Robert Joseph Evans > Attachments: MR-4647.txt, MR-4647.txt, MR-4647.txt > > > For backwards compatibility we recently added made is so we would unjar the > job.jar and add anything to the classpath in the lib directory of that jar. > But this also slows job startup down a lot if the jar is large. We should > only unjar it if actually doing so would add something new to the classpath. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (MAPREDUCE-4659) Confusing output when running "hadoop version" from one hadoop installation when HADOOP_HOME points to another
[ https://issues.apache.org/jira/browse/MAPREDUCE-4659?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Sandy Ryza updated MAPREDUCE-4659: -- Affects Version/s: 1.0.0 > Confusing output when running "hadoop version" from one hadoop installation > when HADOOP_HOME points to another > -- > > Key: MAPREDUCE-4659 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-4659 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: client >Affects Versions: 0.20.2 >Reporter: Sandy Ryza > > Hadoop version X is downloaded to ~/hadoop-x, and Hadoop version Y is > downloaded to ~/hadoop-y. HADOOP_HOME is set to hadoop-x. A user running > hadoop-y/bin/hadoop might expect to be running the hadoop-y jars, but, > because of HADOOP_HOME, will actually be running hadoop-x jars. > "hadoop version" could help clear this up a little by reporting the current > HADOOP_HOME. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Updated] (MAPREDUCE-4659) Confusing output when running "hadoop version" from one hadoop installation when HADOOP_HOME points to another
[ https://issues.apache.org/jira/browse/MAPREDUCE-4659?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] Sandy Ryza updated MAPREDUCE-4659: -- Affects Version/s: (was: 2.0.1-alpha) (was: 1.0.0) 0.20.2 > Confusing output when running "hadoop version" from one hadoop installation > when HADOOP_HOME points to another > -- > > Key: MAPREDUCE-4659 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-4659 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: client >Affects Versions: 0.20.2 >Reporter: Sandy Ryza > > Hadoop version X is downloaded to ~/hadoop-x, and Hadoop version Y is > downloaded to ~/hadoop-y. HADOOP_HOME is set to hadoop-x. A user running > hadoop-y/bin/hadoop might expect to be running the hadoop-y jars, but, > because of HADOOP_HOME, will actually be running hadoop-x jars. > "hadoop version" could help clear this up a little by reporting the current > HADOOP_HOME. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (MAPREDUCE-4644) mapreduce-client-jobclient-tests do not run from dist tarball
[ https://issues.apache.org/jira/browse/MAPREDUCE-4644?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13455994#comment-13455994 ] Alejandro Abdelnur commented on MAPREDUCE-4644: --- Opened MAPREDUCE-4644 > mapreduce-client-jobclient-tests do not run from dist tarball > - > > Key: MAPREDUCE-4644 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-4644 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: build, test >Affects Versions: 2.0.2-alpha >Reporter: Jason Lowe >Priority: Blocker > > The mapreduce jobclient tests rely on junit which is missing from the dist > tarball. This prevents running often-used tests like sleep jobs. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Created] (MAPREDUCE-4659) Confusing output when running "hadoop version" from one hadoop installation when HADOOP_HOME points to another
Sandy Ryza created MAPREDUCE-4659: - Summary: Confusing output when running "hadoop version" from one hadoop installation when HADOOP_HOME points to another Key: MAPREDUCE-4659 URL: https://issues.apache.org/jira/browse/MAPREDUCE-4659 Project: Hadoop Map/Reduce Issue Type: Bug Components: client Affects Versions: 2.0.1-alpha Reporter: Sandy Ryza Hadoop version X is downloaded to ~/hadoop-x, and Hadoop version Y is downloaded to ~/hadoop-y. HADOOP_HOME is set to hadoop-x. A user running hadoop-y/bin/hadoop might expect to be running the hadoop-y jars, but, because of HADOOP_HOME, will actually be running hadoop-x jars. "hadoop version" could help clear this up a little by reporting the current HADOOP_HOME. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (MAPREDUCE-4502) Multi-level aggregation with combining the result of maps per node/rack
[ https://issues.apache.org/jira/browse/MAPREDUCE-4502?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13455909#comment-13455909 ] Chris Douglas commented on MAPREDUCE-4502: -- bq. This seems to be good approach to deal with rack-level aggregation. Do you have any results about the benchmark? For reducing on key ranges, there's a paper in [SOCC|http://www.socc2012.org/papers] on Sailfish. I don't have a link to that paper, though there's a [tech report|http://research.yahoo.com/files/yl-2012-003.pdf]. For the benchmark, we were mostly handling cases without combiners; in our data, each combiner was too effective to benefit from an intermediate level. > Multi-level aggregation with combining the result of maps per node/rack > --- > > Key: MAPREDUCE-4502 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-4502 > Project: Hadoop Map/Reduce > Issue Type: Improvement > Components: applicationmaster, mrv2 >Reporter: Tsuyoshi OZAWA >Assignee: Tsuyoshi OZAWA > Attachments: speculative_draft.pdf > > > The shuffle costs is expensive in Hadoop in spite of the existence of > combiner, because the scope of combining is limited within only one MapTask. > To solve this problem, it's a good way to aggregate the result of maps per > node/rack by launch combiner. > This JIRA is to implement the multi-level aggregation infrastructure, > including combining per container(MAPREDUCE-3902 is related), coordinating > containers by application master without breaking fault tolerance of jobs. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Created] (MAPREDUCE-4658) Move tools JARs into separate lib directories and have common bootstrap script.
Alejandro Abdelnur created MAPREDUCE-4658: - Summary: Move tools JARs into separate lib directories and have common bootstrap script. Key: MAPREDUCE-4658 URL: https://issues.apache.org/jira/browse/MAPREDUCE-4658 Project: Hadoop Map/Reduce Issue Type: Improvement Affects Versions: 2.0.2-alpha Reporter: Alejandro Abdelnur Assignee: Alejandro Abdelnur This is a follow up of the discussion going on on MAPREDUCE-4644 -- Moving each tools JARs into separate lib/ dirs it is quite easy (modifying a single assembly). What we should think is a common bootstrap script for that so each tool does not have to duplicate (and get wrong) such script. I'll open a JIRA for that. -- -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (MAPREDUCE-4644) mapreduce-client-jobclient-tests do not run from dist tarball
[ https://issues.apache.org/jira/browse/MAPREDUCE-4644?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13455900#comment-13455900 ] Alejandro Abdelnur commented on MAPREDUCE-4644: --- Moving each tools JARs into separate lib/ dirs it is quite easy (modifying a single assembly). What we should think is a common bootstrap script for that so each tool does not have to duplicate (and get wrong) such script. I'll open a JIRA for that. > mapreduce-client-jobclient-tests do not run from dist tarball > - > > Key: MAPREDUCE-4644 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-4644 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: build, test >Affects Versions: 2.0.2-alpha >Reporter: Jason Lowe >Priority: Blocker > > The mapreduce jobclient tests rely on junit which is missing from the dist > tarball. This prevents running often-used tests like sleep jobs. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (MAPREDUCE-4651) Benchmarking random reads with DFSIO
[ https://issues.apache.org/jira/browse/MAPREDUCE-4651?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13455895#comment-13455895 ] Ravi Prakash commented on MAPREDUCE-4651: - Thanks Konstantin! I applied the patch and ran the random and backward read tests on my single node dev box. {noformat} $HADOOP_PREFIX/bin/hadoop org.apache.hadoop.fs.TestDFSIO -read -random -fileSize 10MB Average IO rate mb/sec: 134.43310546875 IO rate std deviation: 0.0089636501456 $HADOOP_PREFIX/bin/hadoop org.apache.hadoop.fs.TestDFSIO -read -backward -fileSize 10MB Average IO rate mb/sec: 134.49253845214844 IO rate std deviation: 0.026679629420752023 $HADOOP_PREFIX/bin/hadoop org.apache.hadoop.fs.TestDFSIO -read -random -fileSize 1GB Average IO rate mb/sec: 249.47183227539062 IO rate std deviation: 0.014617091655162118 $HADOOP_PREFIX/bin/hadoop org.apache.hadoop.fs.TestDFSIO -read -backward -fileSize 1GB Average IO rate mb/sec: 295.8538818359375 IO rate std deviation: 0.061419808441541615 $HADOOP_PREFIX/bin/hadoop org.apache.hadoop.fs.TestDFSIO -read -random -fileSize 10GB Average IO rate mb/sec: 320.3417663574219 IO rate std deviation: 0.05935480659067817 $HADOOP_PREFIX/bin/hadoop org.apache.hadoop.fs.TestDFSIO -read -backward -fileSize 10GB Average IO rate mb/sec: 323.28045654296875 IO rate std deviation: 0.0598550775330073 $HADOOP_PREFIX/bin/hadoop org.apache.hadoop.fs.TestDFSIO -read -backward -fileSize 30GB Average IO rate mb/sec: 390.9880065917969 IO rate std deviation: 0.06083891027478396 $HADOOP_PREFIX/bin/hadoop org.apache.hadoop.fs.TestDFSIO -read -random -fileSize 30GB Average IO rate mb/sec: 369.2136535644531 IO rate std deviation: 0.056819116587427144 {noformat} Could you please post recommended usage? And at what sizes do we expect to achieve stable IO rates? > Benchmarking random reads with DFSIO > > > Key: MAPREDUCE-4651 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-4651 > Project: Hadoop Map/Reduce > Issue Type: New Feature > Components: benchmarks, test >Affects Versions: 1.0.0 >Reporter: Konstantin Shvachko >Assignee: Konstantin Shvachko > Attachments: randomDFSIO.patch, randomDFSIO.patch > > > TestDFSIO measures throughput of HDFS write, read, and append operations. It > will be useful to have an option to use it for benchmarking random reads. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (MAPREDUCE-4502) Multi-level aggregation with combining the result of maps per node/rack
[ https://issues.apache.org/jira/browse/MAPREDUCE-4502?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13455842#comment-13455842 ] Tsuyoshi OZAWA commented on MAPREDUCE-4502: --- s/Do you have some result to/Do you have any results about the benchmark?/ > Multi-level aggregation with combining the result of maps per node/rack > --- > > Key: MAPREDUCE-4502 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-4502 > Project: Hadoop Map/Reduce > Issue Type: Improvement > Components: applicationmaster, mrv2 >Reporter: Tsuyoshi OZAWA >Assignee: Tsuyoshi OZAWA > Attachments: speculative_draft.pdf > > > The shuffle costs is expensive in Hadoop in spite of the existence of > combiner, because the scope of combining is limited within only one MapTask. > To solve this problem, it's a good way to aggregate the result of maps per > node/rack by launch combiner. > This JIRA is to implement the multi-level aggregation infrastructure, > including combining per container(MAPREDUCE-3902 is related), coordinating > containers by application master without breaking fault tolerance of jobs. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (MAPREDUCE-4502) Multi-level aggregation with combining the result of maps per node/rack
[ https://issues.apache.org/jira/browse/MAPREDUCE-4502?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13455840#comment-13455840 ] Tsuyoshi OZAWA commented on MAPREDUCE-4502: --- Chris and Karthik, Thank you for your sharing your experience and thinking. These are very useful for me. bq. ShuffleHandler is an auxiliary service loaded in the NodeManager. It's shared across all containers. I see. I have to redesign it to run combiner in container. bq. Carlo Curino and I experimented with this, but (a) saw only slight improvements in job performance and (b) the changes to the AM to accommodate a new task type were extensive. This is very interesting. In fact, I prototyped to run combiner at the end of MapTask as the first version. And, its performance was good. In this case, I found that it's needed to add new status to MapTask because of assuring fault tolerance. Is it acceptable for hadoop to do that? bq. With logic to manage skew, we're hoping that scheduling an aggressive range can have a similar effect to combiner tasks, without introducing the new task type. This seems to be good approach to deal with rack-level aggregation. Do you have some result to bq. 1. Perform node-level aggregation (reduce) at the end of maps in co-ordination with AM. bq. 2. Perform rack-level aggregation at the end of node-level aggregation again in co-ordination with AM. The aggregation could be performed in parallel across the involved nodes such that each node has aggregated values of different keys. bq. 3. Schedule reducers taking the key-distribution into account across racks. Nice wrap-up :-) bq. The con will be that the shuffle won't be asynchronous to map computation, but hopefully this wouldn't offset the gains of decreased network and disk I/O. The balance between the gains by asynchronous processing and the one by decreasing network and disk I/O. In my previous experiment, it deeply depends on number of reducers. I think these gains are trade-off, so parameters are necessary to deal with various workloads. bq. PS. http://dl.acm.org/citation.cfm?id=1901088 documents the advantages of multi-level aggregation in the context of graph algorithms modeled as iterative MR jobs. I'm going read it :) It's time for me to create the new revision of design note with reflecting your opinion. Thanks, -- Tsuyoshi > Multi-level aggregation with combining the result of maps per node/rack > --- > > Key: MAPREDUCE-4502 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-4502 > Project: Hadoop Map/Reduce > Issue Type: Improvement > Components: applicationmaster, mrv2 >Reporter: Tsuyoshi OZAWA >Assignee: Tsuyoshi OZAWA > Attachments: speculative_draft.pdf > > > The shuffle costs is expensive in Hadoop in spite of the existence of > combiner, because the scope of combining is limited within only one MapTask. > To solve this problem, it's a good way to aggregate the result of maps per > node/rack by launch combiner. > This JIRA is to implement the multi-level aggregation infrastructure, > including combining per container(MAPREDUCE-3902 is related), coordinating > containers by application master without breaking fault tolerance of jobs. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (MAPREDUCE-4644) mapreduce-client-jobclient-tests do not run from dist tarball
[ https://issues.apache.org/jira/browse/MAPREDUCE-4644?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13455826#comment-13455826 ] Robert Joseph Evans commented on MAPREDUCE-4644: We need to be in the process of separating out true unit test from system and integration tests. Unit tests should run fast and are something that we can do for all components as part of the pre-commit build. If the test is full featured enough that it could be called as a "tool" then there is no way that it is a true unit test. I am +1 for moving those out to be part of a tools or examples package somewhere. We also need to look at cleaning up our classpaths. Separating each tool out into a directory with a full list of its dependencies seems like a reasonable solution. It is what Oozie asks users to do for their work flows and seems to work fairly well. But that starts to sound like a larger effort then moving a few classes around and splitting a launcher into two. I think it is something that needs to be done, but perhaps needs some design work, especially in relation to how we may want to do dependency isolation in the future with OSGi or something else. Alejandro, I know you have been looking at and thinking about the classpath issue with YARN/MR, and how we should package things a lot already. MAPREDUCE-3745, HADOOP-7935, and MAPREDUCE-4421 do we need another JIRA explicitly for tools? How do we handle the case of a tool having a map/reduce dependency now that the MR code is going to be separated out so that we can use a different version of MR? Does that mean that they have to provide their own tools with a MR dependency and a config to point to them? It just seems like a change like this needs a full design. > mapreduce-client-jobclient-tests do not run from dist tarball > - > > Key: MAPREDUCE-4644 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-4644 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: build, test >Affects Versions: 2.0.2-alpha >Reporter: Jason Lowe >Priority: Blocker > > The mapreduce jobclient tests rely on junit which is missing from the dist > tarball. This prevents running often-used tests like sleep jobs. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (MAPREDUCE-3357) TestMRWithDistributedCache fails on branch-20-security
[ https://issues.apache.org/jira/browse/MAPREDUCE-3357?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13455821#comment-13455821 ] Amir Sanjar commented on MAPREDUCE-3357: add "chmod a+x $HOME" .. it should fix the problem. let me know > TestMRWithDistributedCache fails on branch-20-security > -- > > Key: MAPREDUCE-3357 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-3357 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: distributed-cache, test >Affects Versions: 0.20.205.0 >Reporter: Eli Collins > > TestMRWithDistributedCache testLocalJobRunner fails on branch-20-security: > {noformat} > Testcase: testLocalJobRunner took 5.501 sec > FAILED > null > junit.framework.AssertionFailedError: null > at > org.apache.hadoop.filecache.TestMRWithDistributedCache.testWithConf(TestMRWithDistributedCache.java:154) > at > org.apache.hadoop.filecache.TestMRWithDistributedCache.testLocalJobRunner(TestMRWithDistributedCache.java:162) > {noformat} -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (MAPREDUCE-4644) mapreduce-client-jobclient-tests do not run from dist tarball
[ https://issues.apache.org/jira/browse/MAPREDUCE-4644?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13455793#comment-13455793 ] Tom White commented on MAPREDUCE-4644: -- bq. If desired we could separate these into "tests that are really tools" which would go into tools/ and shouldn't rely on junit or other test framework stuff and "tests that are really unit tests" that go into something like tests/. +1. Many of the "tests that are really tools" are benchmarks so we could call them that. > mapreduce-client-jobclient-tests do not run from dist tarball > - > > Key: MAPREDUCE-4644 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-4644 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: build, test >Affects Versions: 2.0.2-alpha >Reporter: Jason Lowe >Priority: Blocker > > The mapreduce jobclient tests rely on junit which is missing from the dist > tarball. This prevents running often-used tests like sleep jobs. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira
[jira] [Commented] (MAPREDUCE-4644) mapreduce-client-jobclient-tests do not run from dist tarball
[ https://issues.apache.org/jira/browse/MAPREDUCE-4644?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13455782#comment-13455782 ] Jason Lowe commented on MAPREDUCE-4644: --- Actually I'm thinking of cases where the test jars themselves cause the problems, see HDFS-3831. There are a lot of things in these tests jars besides the items that are invoked by ToolRunner, and not all test jars even use ToolRunner. If desired we could separate these into "tests that are really tools" which would go into tools/ and shouldn't rely on junit or other test framework stuff and "tests that are really unit tests" that go into something like tests/. That would make running the "tests that are tools" a bit easier since we hopefully don't need a separate classpath beyond TOOL_PATH, but the junit test cases are completely out of the way. > mapreduce-client-jobclient-tests do not run from dist tarball > - > > Key: MAPREDUCE-4644 > URL: https://issues.apache.org/jira/browse/MAPREDUCE-4644 > Project: Hadoop Map/Reduce > Issue Type: Bug > Components: build, test >Affects Versions: 2.0.2-alpha >Reporter: Jason Lowe >Priority: Blocker > > The mapreduce jobclient tests rely on junit which is missing from the dist > tarball. This prevents running often-used tests like sleep jobs. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira