Client should be able to know why an AM crashed.
------------------------------------------------
Key: MAPREDUCE-2717
URL: https://issues.apache.org/jira/browse/MAPREDUCE-2717
Project: Hadoop Map/Reduce
Issue Type: Improvement
Components: mrv2
Reporter: Amol Kekre
Fix For: 0.23.0
Today if an AM crashes, we have to dig through logs - very cumbersome. It is
good to have client print some reason for
AM crash. Various possible reasons for AM crash:
(1) AM container failed during localization itself.
(2) AM container launched but failed before properly starting, for e.g. due to
classpath issues
(3) AM failed after starting properly.
(4) an AM is expired and killed by the RM
Potential fixes:
- For (1) and (2) the client should obtain the container-status, container
diagnostics and exit code.
- For (3), the AM should set some kind of reason for failure during its
heartbeat to RM and the client should obtain
the same from RM.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
