Retrying Tasks
Changed in version 8.0.0.
Tasks that fail but are configured to retry return to the waiting
state, with a new clock trigger to handle the configured retry delay.
Note
A task that is waiting on a retry will already have one or more failed jobs associated with it.
Aborting a Retry Sequence
To prevent a waiting task from retrying, remove it from the scheduler’s
active window. For a task 3/foo
in workflow brew
:
$ cylc remove brew//3/foo
If you kill a running task that has more retries configured, it goes to the
held
state so you can decide whether to release it and continue the retry
sequence, or remove it.
$ cylc kill brew//3/foo # 3/foo goes to held state post kill
$ cylc release brew//3/foo # release to continue retrying...
$ cylc remove brew//3/foo # ... OR remove the task to stop retries
If you want trigger downstream tasks despite 3/foo
being removed before it
could succeed, use cylc set-outputs
to artificially mark its succeeded
output as complete (and with the --flow
option, to make the flow
continue on from there).