Opened 6 months ago

Closed 5 months ago

#3368 closed help (answered)

stuck UM suites on ARCHER; login7->login

Reported by: pmcguire Owned by: um_support
Component: UM Model Keywords: UM, login7, login, ARCHER, ssh
Cc: Platform: ARCHER
UM Version: 11.5


Hi CMS Helpdesk:
I was running 3 different N216e UM suites on ARCHER, when we were encouraged to switch from using login7 and to start using login. I made that switch in the suite, and I reloaded the suite. After reloading the suite (i.e., u-bw963), the postproc and pptransfer jobs for cycle 20020801 were successfully submitted and completed through the new login node. But the atmos_main task for the next cycle was submitted on September 12, 11:09 UT, but it hasn't started running yet. This is a lot longer of a queueing time than recently before.

Any suggestions for what I might have done wrong so that the job isn't starting after 3 days?

Change History (6)

comment:1 Changed 6 months ago by grenville


Those jobs are in "H" state - I don't know why. Try stopping the tasks and re-triggering them.


comment:2 Changed 6 months ago by pmcguire

Hi Grenville:
Thanks. I will try that.
I didn't see any H state or Hold state in the Cylc GUI. Is there some other way that you saw that state?

comment:3 Changed 5 months ago by grenville

run qstat on archer

comment:4 Changed 5 months ago by pmcguire

Thanks. That qstat is helpful.
Shouldn't the H state (I guess it means Hold state) also be visible after polling in the Cylc GUI on PUMA?
I already retriggered, so I can't easily test this for myself, since qstat on ARCHER has the jobs in the Q state now.

comment:5 Changed 5 months ago by ros

Hi Patrick,

No, cylc does not know anything about the PBS "hold" queue state and indeed other queue states. It only see the job as queued on ARCHER.


comment:6 Changed 5 months ago by pmcguire

  • Resolution set to answered
  • Status changed from new to closed
Note: See TracTickets for help on using tickets.