exec-dazzling-rose-gecko-of-perception was at 6057/6058 completed invert_field (subchunkable) tasks this morning, and the workers failed due to the missing outcome queue.
Edit: Scheduler was frozen and did not react to Ctrl+C. Had to kill it.
What should happen in this case: The run should fail immediately when the pods keep dying - no need to waste resources with an unrecoverable error.
What also should happen: Not deleting the outcome queue before the last task is actually processed, but I can't replicate it.
exec-dazzling-rose-gecko-of-perception
was at 6057/6058 completedinvert_field
(subchunkable) tasks this morning, and the workers failed due to the missing outcome queue. Edit: Scheduler was frozen and did not react to Ctrl+C. Had to kill it.What should happen in this case: The run should fail immediately when the pods keep dying - no need to waste resources with an unrecoverable error.
What also should happen: Not deleting the outcome queue before the last task is actually processed, but I can't replicate it.