Failing workflow leaves workflow and history behind

[anilthanki]

I am trying this with example workflow provided in repo, its failing with the following error and it does not delete history and workflow in Galaxy instance

Failure when invoking invoke workflows: Unexpected HTTP status code: 400: {"err_msg": {"header": "No value found for 'Dataset has a header'. Using default: 'False'."}, "err_code": 0, "err_data": {"3": {"header": "No value found for 'Dataset has a header'. Using default: 'False'."}}}

[pcm32]

That should probably be optional (keeping the default as it is currently). You want to keep the workflow and history for debugging purposes, but you might as well want them removed. So maybe you can add a flag to enable that new behaviour.

[pcm32]

It might be that the tool installed on that instance is newer and has an option that is unaccounted for in the execution of the workflow (either workflow file or parameters).

[pcm32]

I would suggest opening the workflow in the UI, it will update any missing tools, save it and then obtain the workflow file again from there. Then I would try with that one.

[pcm32]

Also note that the communication process could also fail, but the workflow keeps running on Galaxy. We have a mechanism by which on the next same run it will first check if this is the case (for which the workflow and the history need to remain as well) and just keep tracking the running workflow. So what I'm trying to say is that adding that functionality should also take into account if the failure is at the level of galaxy running the pipeline or a communication failure (in which case you also don't want to delete anything).