Bug #16663
Updated by Ward Vandewege over 4 years ago
Restarting a-d-c seems to have rather nasty consequences for running workflows.
1) All running jobs are aborted immediately (that's bad).
2) The CR then gets associated with a new container, which automatically restarts all jobs that are runnable and haven't run yet (I think). But, it's suddenly impossible to see the container that was associated with the CR originally, which means that it looks like all work is just gone, lost (that's also bad, but mostly a UX problem).
The old container is still there, and whatever was run under that will be reused if the jobs were not marked as non-deterministic (I think).
Example:
CR: pirca-xvhdp-dsi31r62tcg7lxi
original container: pirca-dz642-k1o7j58y4eholt0
new container: pirca-dz642-a4vyxbsz8himlpc
Question: a) is this expected or just a (nasty) bug? b) regardless of the answer to a, how do we make a-d-c not a thing that costs tons of money (and creates a lot of user confusion) when it dies?