I see how the story description and API docs imply otherwise, but non-admin users are already prohibited from updating containers at all (except the container process itself can update its own progress fields while state==Running). So I don't think we need a special "admin state transitions" here.
In docs:
For example, if a container exited successfully but produced bad output, it may not be worth updating the workflow to use a new image/version. Instead, changing the state of a container to Cancelled
will disable reuse of the specific container.
I think this should be presented as an expedient thing to do in addition to fixing the bug, rather than as an alternative. We can't force people to fix the bugs, but I don't think we should nudge them in the wrong direction.
Maybe: "... it may not be feasible to update the workflow immediately ... Meanwhile, changing the state ..."
...will change it's state to...
its
...where -u
is the UUID
of the container:
...where xxxxx-xxxxx-xxxxxxxxxxxxxxx
is the UUID of the container
all containers that have been cancelled after completion
Either this should say "exited 0 and were then cancelled" or the condition should be ["exit_code","<>",null]
\ No newline at end of file
(manage-containers.html.textile.liquid) should have one.
Cancelled (admin only)
All state changes are admin only.
The admin page title should probably be something like "Controlling container reuse" -- hint at the user's question, which they know, rather than the answer, which they don't. Same goes for the leading sentence. And I think it needs to be added to _config.yml so it shows up in the left nav.