Details
-
Bug
-
Status: Resolved
-
Minor
-
Resolution: Incomplete
-
2.3.1
-
None
Description
I met a issue that if I start a spark application use yarn client mode, application sometimes hang.
I check the application logs, container allocate on a lost NodeManager, but AM don't retry to start another executor.
My spark version is 2.3.1
Here is my ApplicationMaster log.
2018-09-26 05:21:15 INFO YarnRMClient:54 - Registering the ApplicationMaster
2018-09-26 05:21:15 INFO ConfiguredRMFailoverProxyProvider:100 - Failing over to rm2
2018-09-26 05:21:15 WARN Utils:66 - spark.executor.instances less than spark.dynamicAllocation.minExecutors is invalid, ignoring its setting, please update your configs.
2018-09-26 05:21:15 INFO Utils:54 - Using initial executors = 1, max of spark.dynamicAllocation.initialExecutors, spark.dynamicAllocation.minExecutors and spark.executor.instances
2018-09-26 05:21:15 INFO YarnAllocator:54 - Will request 1 executor container(s), each with 24 core(s) and 20275 MB memory (including 1843 MB of overhead)
2018-09-26 05:21:15 INFO YarnAllocator:54 - Submitted 1 unlocalized container requests.
2018-09-26 05:21:15 INFO ApplicationMaster:54 - Started progress reporter thread with (heartbeat : 3000, initial allocation : 200) intervals
2018-09-26 05:21:27 WARN YarnAllocator:66 - Cannot find executorId for container: container_1532951609168_4721728_01_000002
2018-09-26 05:21:27 INFO YarnAllocator:54 - Completed container container_1532951609168_4721728_01_000002 (state: COMPLETE, exit status: -100)
2018-09-26 05:21:27 WARN YarnAllocator:66 - Container marked as failed: container_1532951609168_4721728_01_000002. Exit status: -100. Diagnostics: Container released on a lost node