Bug #12916
closed[Documentation] container_requests methods
Description
For container_requests, I'm referring to the page: https://doc.arvados.org/api/methods/container_requests.html
For container_requests, field 'output_uuid' states 'Null if the container has failed or not 'yet completed. However, I can see that failed runs still have this uuid, they're not null. For example, test run https://workbench.e51c5.arvadosapi.com/container_requests/e51c5-xvhdp-edcfrqhr25s6oy7
Updated by Tom Clegg about 7 years ago
I'm inclined to say the documentation is what we want, and the actual behavior is a bug. "Wait for the CR to be final, then use output" seems like a reasonable happy path, compared to explicitly looking up the container and checking its exit_code just to find out whether the output is valid.
One caveat is that the partial/broken output collection has to be created (in order for the user to have read permission) and this makes it a bit more troublesome to tidy up old outputs. Should we set a trash time on partial/broken outputs even when output_ttl isn't used?
12916-no-cr-output-on-fail @ c7a6ab35c1537d2766bdaaa7fec83a15c5454094
Updated by Tom Clegg about 7 years ago
- Related to Bug #12891: [crunch2] log collection not saved for cancelled job added