a suite has stopped a few times in the last month or so.
investigation into cylc-run/SUITE/log/job shows:
20201118T1500Z/TASK/01/job.err
PuTTY X11 proxy: Authorisation not recognised
PuTTY X11 proxy: Authorisation not recognised
NOTICE: Holland B parameter less than 1; resetting to 1.
NOTE: 26 valid data records were written to ‘MYFILE’.
PuTTY X11 proxy: Authorisation not recognised
PuTTY X11 proxy: Authorisation not recognised
this is most likely not a ‘cylc’ problem.
we have some accounts such that more than one user can run in the account.
it looks someone is using a Windows box and that some environment settings for that user are causing this?
can someone shed light on the interaction so that i can track down this error.
BTW. the suite is running on a Linux cluster.
michael
Hi Michael,
I see your error messages are in a task job log, not the scheduler log. So by “a suite has stopped” I presume you mean that the workflow stalled because the job unexpectedly failed, not that the scheduler experienced an error and shut down?
If so, then I agree that this is unlikely to be a Cylc problem. Cylc wraps your job scripting in straightforward bash code to send a job status message (via cylc message) to the scheduler if kill or exit signals are detected, but there is no Cylc job process listening on the network - only the scheduler does that. So it looks to me as if PuTTY is interfering in the user-defined job code somehow.
If it somehow fatally interfered with the short-lived cylc message process you should see evidence in the job error log that the message could not be sent, and the task would appear stuck as running in the scheduler (because the status change message did not get through).
Regards,
Hilary