Details
-
Bug
-
Status: Closed
-
Major
-
Resolution: Fixed
-
None
-
None
-
None
Description
Currently, if not queue or a custom queue is set for an action, the launcher job will use the same queue.
This leads to a deadlock situation in small clusters where all slots are taken by launcher jobs that cannot progress because there are not slots avail.
The solution is to allow launchers to use a different queue (i.e. oozie.launcher.mapred.job.queue.name=launcherqueue).
This can be done today, but if the action defines a custom queue (i.e. mapred.job.queue.name=fooqueue) the above property is overwritten and both launcher and job end up in the same queue.
The JavaActionExecutor should do a special check.