Closed matthewrmshin closed 10 years ago
@matthewrmshin - no we're not relying on the job script error trap. If the kill command succeeds, the poll-and-kill thread puts "kill command succeeded" in the task message queue, and that should be picked up in task.py and result in a task state change. I'll see if I can reproduce the problem.
It works for me - I used an LL directive output = /illegal
for an unwritable job output file to make LL put the queued job on "User Hold" indefinitely, then I attempted to kill it with cylc. This cancelled the job and cylc put the task in the submit-failed state. Must be something odd happening at your end?
@matthewrmshin - can we close this?
I need to have another look at this.
This is no longer an issue. (Tested using latest master.)
Killing a loadleveler job at the submitted state (via
cylc kill
orgcylc
) does not appear to update the status of the suite.Is it because we are relying on the trap of the job script to send a message back to the suite to tell the suite that the job has been killed? (Obviously, the trap would not work if the job script is not even started.)