Reputation: 60014
I just discovered that killing a hive process does not kill the corresponding hadoop jobs.
I had to issue the hadoop job -kill
command and drop the table by hand.
Is there a way to automate this: i.e., if a hive job is killed, all the hadoop jobs it spawned should be killed too (and the intermediate tables dropped)?
Upvotes: 0
Views: 360
Reputation: 63082
I am not aware of such a mechanism and have been doing precisely as you mention to do hadoop job -kill on the M/R job. That should only be required of the current stage of the hive job: the succeeding stages would not be launched.
Upvotes: 0