spark-instrumented-optimizer/resource-managers/yarn
Yuming Wang 04e71c3160 [MINOR][YARN] Add disable yarn.nodemanager.vmem-check-enabled option to memLimitExceededLogMessage
My spark application sometimes will throw `Container killed by YARN for exceeding memory limits`.
Even I increased `spark.yarn.executor.memoryOverhead` to 10G, this error still happen.  The latest config:
<img width="685" alt="memory-config" src="https://user-images.githubusercontent.com/5399861/36975716-f5c548d2-20b5-11e8-95e5-b228d50917b9.png">

And error message:
```
ExecutorLostFailure (executor 121 exited caused by one of the running tasks) Reason: Container killed by YARN for exceeding memory limits. 30.7 GB of 30 GB physical memory used. Consider boosting spark.yarn.executor.memoryOverhead.
```

This is because of [Linux glibc >= 2.10 (RHEL 6) malloc may show excessive virtual memory usage](https://www.ibm.com/developerworks/community/blogs/kevgrig/entry/linux_glibc_2_10_rhel_6_malloc_may_show_excessive_virtual_memory_usage?lang=en). So disable `yarn.nodemanager.vmem-check-enabled` looks like a good option as [MapR mentioned ](https://mapr.com/blog/best-practices-yarn-resource-management).

This PR add disable `yarn.nodemanager.vmem-check-enabled` option to memLimitExceededLogMessage.

More details:
https://issues.apache.org/jira/browse/YARN-4714
https://stackoverflow.com/a/31450291
https://stackoverflow.com/a/42091255

After this PR:
<img width="898" alt="yarn" src="https://user-images.githubusercontent.com/5399861/36975949-c8e7bbbe-20b6-11e8-9513-9f903b868d8d.png">

N/A

Author: Yuming Wang <yumwang@ebay.com>
Author: Yuming Wang <wgyumg@gmail.com>

Closes #20735 from wangyum/YARN-4714.

Change-Id: Ie10836e2c07b6384d228c3f9e89f802823bd9f16
2018-03-07 17:01:29 +08:00
..
src [MINOR][YARN] Add disable yarn.nodemanager.vmem-check-enabled option to memLimitExceededLogMessage 2018-03-07 17:01:29 +08:00
pom.xml [SPARK-23028] Bump master branch version to 2.4.0-SNAPSHOT 2018-01-13 00:37:59 +08:00