You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The default action when performing a restart should be to start the suite in exactly the same state defined in the restart file. This is the safest behaviour, is the least likely to surprise the user and is necessary if we want cylc to be able update the status of tasks upon restart (see #86). However, this is not the behaviour at present. Even using the options --no-reset and --no-release you still end up with, for instance, any tasks which were in the submitted or running states being re-run.
The text was updated successfully, but these errors were encountered:
Current behaviour was intended to save the operator, potentially, from a lot of manual triggering - imagine loadleveler going down just after a large ensemble is submitted by cylc - and it is "safe" in the sense that it assumes the worst (any task that was not finished at shutdown must be re-run) and if a failed task was not fixed it will just fail again so we're no worse off.
That said, your solution provides the ultimate safety and avoids unnecessary re-runs, and we could keep the current behaviour as an option (e.g. 'cylc restart --trigger-all') and/or provide run time controls to re-trigger multiple tasks at once. In other words, I agree.
The default action when performing a restart should be to start the suite in exactly the same state defined in the restart file. This is the safest behaviour, is the least likely to surprise the user and is necessary if we want cylc to be able update the status of tasks upon restart (see #86). However, this is not the behaviour at present. Even using the options --no-reset and --no-release you still end up with, for instance, any tasks which were in the submitted or running states being re-run.
The text was updated successfully, but these errors were encountered: