[
https://issues.apache.org/jira/browse/MAPREDUCE-7435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17731016#comment-17731016
]
ASF GitHub Bot commented on MAPREDUCE-7435:
-------------------------------------------
steveloughran opened a new pull request, #5728:
URL: https://github.com/apache/hadoop/pull/5728
This modifies the manifest committer so that the list of files to rename is
passed between stages as a file of writeable entries on the local filesystem.
The map of directories to create is still passed in memory; this map is
built across all tasks, so even if many tasks created files, if they all write
into the same set of directories the memory needed is O(directories) with the
task count not a factor.
The _SUCCESS file reports on heap size through gauges. This should give a
warning if there are problems.
Contributed by Steve Loughran
### How was this patch tested?
Itests in progress; yetus gets to test mapreduce
### For code changes:
- [ ] Does the title or this PR starts with the corresponding JIRA issue id
(e.g. 'HADOOP-17799. Your PR title ...')?
- [ ] Object storage: have the integration tests been executed and the
endpoint declared according to the connector-specific documentation?
- [ ] If adding new dependencies to the code, are these dependencies
licensed in a way that is compatible for inclusion under [ASF
2.0](http://www.apache.org/legal/resolved.html#category-a)?
- [ ] If applicable, have you updated the `LICENSE`, `LICENSE-binary`,
`NOTICE-binary` files?
> ManifestCommitter OOM on azure job
> ----------------------------------
>
> Key: MAPREDUCE-7435
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-7435
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Components: client
> Affects Versions: 3.3.5
> Reporter: Steve Loughran
> Assignee: Steve Loughran
> Priority: Major
> Labels: pull-request-available
>
> I've got some reports of spark jobs OOM if the manifest committer is used
> through abfs.
> either the manifests are using too much memory, or something is not working
> with azure stream memory use (or both).
> before proposing a solution, first step should be to write a test to load
> many, many manifests, each with lots of dirs and files to see what breaks.
> note: we did have OOM issues with the s3a committer, on teragen but those
> structures have to include every etag of every block, so the manifest size is
> O(blocks); the new committer is O(files + dirs).
> {code}
> java.lang.OutOfMemoryError: Java heap space
> at
> org.apache.hadoop.fs.azurebfs.services.AbfsInputStream.readOneBlock(AbfsInputStream.java:314)
> at
> org.apache.hadoop.fs.azurebfs.services.AbfsInputStream.read(AbfsInputStream.java:267)
> at java.io.DataInputStream.read(DataInputStream.java:149)
> at
> com.fasterxml.jackson.core.json.ByteSourceJsonBootstrapper.ensureLoaded(ByteSourceJsonBootstrapper.java:539)
> at
> com.fasterxml.jackson.core.json.ByteSourceJsonBootstrapper.detectEncoding(ByteSourceJsonBootstrapper.java:133)
> at
> com.fasterxml.jackson.core.json.ByteSourceJsonBootstrapper.constructParser(ByteSourceJsonBootstrapper.java:256)
> at com.fasterxml.jackson.core.JsonFactory._createParser(JsonFactory.java:1656)
> at com.fasterxml.jackson.core.JsonFactory.createParser(JsonFactory.java:1085)
> at
> com.fasterxml.jackson.databind.ObjectMapper.readValue(ObjectMapper.java:3585)
> at
> org.apache.hadoop.util.JsonSerialization.fromJsonStream(JsonSerialization.java:164)
> at org.apache.hadoop.util.JsonSerialization.load(JsonSerialization.java:279)
> at
> org.apache.hadoop.mapreduce.lib.output.committer.manifest.files.TaskManifest.load(TaskManifest.java:361)
> at
> org.apache.hadoop.mapreduce.lib.output.committer.manifest.impl.ManifestStoreOperationsThroughFileSystem.loadTaskManifest(ManifestStoreOperationsThroughFileSystem.java:133)
> at
> org.apache.hadoop.mapreduce.lib.output.committer.manifest.stages.AbstractJobOrTaskStage.lambda$loadManifest$6(AbstractJobOrTaskStage.java:493)
> at
> org.apache.hadoop.mapreduce.lib.output.committer.manifest.stages.AbstractJobOrTaskStage$$Lambda$231/1813048085.apply(Unknown
> Source)
> at
> org.apache.hadoop.fs.statistics.impl.IOStatisticsBinding.invokeTrackingDuration(IOStatisticsBinding.java:543)
> at
> org.apache.hadoop.fs.statistics.impl.IOStatisticsBinding.lambda$trackDurationOfOperation$5(IOStatisticsBinding.java:524)
> at
> org.apache.hadoop.fs.statistics.impl.IOStatisticsBinding$$Lambda$217/489150849.apply(Unknown
> Source)
> at
> org.apache.hadoop.fs.statistics.impl.IOStatisticsBinding.trackDuration(IOStatisticsBinding.java:445)
> at
> org.apache.hadoop.mapreduce.lib.output.committer.manifest.stages.AbstractJobOrTaskStage.loadManifest(AbstractJobOrTaskStage.java:492)
> at
> org.apache.hadoop.mapreduce.lib.output.committer.manifest.stages.LoadManifestsStage.fetchTaskManifest(LoadManifestsStage.java:170)
> at
> org.apache.hadoop.mapreduce.lib.output.committer.manifest.stages.LoadManifestsStage.processOneManifest(LoadManifestsStage.java:138)
> at
> org.apache.hadoop.mapreduce.lib.output.committer.manifest.stages.LoadManifestsStage$$Lambda$229/137752948.run(Unknown
> Source)
> at
> org.apache.hadoop.util.functional.TaskPool$Builder.lambda$runParallel$0(TaskPool.java:410)
> at
> org.apache.hadoop.util.functional.TaskPool$Builder$$Lambda$230/467893357.run(Unknown
> Source)
> at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511)
> at java.util.concurrent.FutureTask.run(FutureTask.java:266)
> at
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
> at java.lang.Thread.run(Thread.java:750)
> {code}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: mapreduce-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: mapreduce-issues-h...@hadoop.apache.org
