Hello! If a workflow fails on a certain step, after fixing whatever issue causing it, what’s the best way to resume that step and let the workflow continue? I don’t mean to rerun that step (like using the rerun button on the UI), that will trigger a new invocation. We want to continue the original workflow invocation by restarting from the failed step and continue to the end of workflow, and have the result still associated with the original invocation. Thanks!
1 Like
When you go to re-run the step, it will give you an option to replace the dataset in the workflow and resume all dependent steps.
1 Like
Hello! Thanks for the tip. I found the option you mentioned, that’s very helpful!
On a follow up question, we actually have a workflow invocation with one step which actually successfully produced the outputs, but the step failed while job runner wrote info to the Galaxy log (some random I/O). We made a fix on this to catch such log I/O error, and allow step to succeed so long as valid outputs are produced. Would you be interested in a PR for this?
For the failed step: since that is an expensive step involving human labor, we want to just update the step status to succeed, and allow the depending steps to continue. I see There is an API for this: PUT /api/workflows/{workflow_id}/invocations/{invocation_id}/steps/{step_id}. However, I don’t know what to put in payload, especially the action. Could you give me an example?