hey,
i can’t seem to restart a cylc
8 workflow.
when i run cylc play WORKFLOW/run1
the cylc log
command gives me…
2023-04-19T05:25:07Z WARNING - No active tasks matching: */*
2023-04-19T05:25:31Z INFO - [command] resume
2023-04-19T05:25:32Z INFO - RESUMING the workflow now
2023-04-19T05:25:32Z INFO - Command succeeded: resume()
2023-04-19T05:25:32Z INFO - Workflow shutting down - AUTOMATIC
2023-04-19T05:25:33Z INFO - [('workflow-event-handler', 'shutdown') cmd]
mail -s '[workflow shutdown] u-cv956/run1' -r notifications@w-cylc02.maui.niwa.co.nz williamsjh <<<'event: shutdown
message: AUTOMATIC
workflow: u-cv956/run1
host: w-cylc02.maui.niwa.co.nz
port: 43057
owner: williamsjh
'
[('workflow-event-handler', 'shutdown') ret_code] 0
2023-04-19T05:25:33Z INFO - DONE
any idea what’s going on? it has several years still to run!
thanks for any ideas!
j
The AUTOMATIC shutdown message means that there is nothing left to run, according to the workflow state recorded in the run DB, i.e. the scheduler has determined that it already finished
How did it shut down exactly, before you tried the restart? Eg. did it shut down automatically as finished then, or did you tell it to stop early?
You should be able to find the scheduler log from that run, in the log/scheduler/
directory.
Whatever the cause, you can get it to carry on by restarting with cylc play --pause
, then triggering the next tasks before un-pausing. But it would be good to check that some kind of erroneous early stop did not occur.
(Note, in a near-future release the scheduler will automatically stall on a timeout to allow intervention, without needing --pause
, if you restart an already-finished workflow).
1 Like
We have a new issue up on the repository, to make shutdown reasons clearer in the logs.
1 Like
cool thanks,
using cylc play --pause
allowed me to trigger the tasks and it’s running again now.
one thing to note (and bearing in mind that this particular run has been monumentally messed around with by me) is that it seems to be running everything in the none
flow! i don’t know precisely how this has happened but it might account for the scheduler not thinking there was anything else to run.
cylc log
output…
cheers
OK, you may find that the it stops again when those tasks have completed!
The “flow=none” is not a flow - i.e. those tasks won’t spawn any action downstream in the graph.
In fact this explains why your workflow automatically shut down as finished - you had only (re)triggered tasks with --flow=none
.
If you want it to carry on at this point, you actually need to trigger a new flow at the appropriate point in the graph.
1 Like
thanks! yeah that all makes sense and is done now. from a previous comment…
in a near-future release the scheduler will automatically stall on a timeout to allow intervention, without needing --pause
, if you restart an already-finished workflow…
yeah this would be ace.
cheers,
j
1 Like