search

lanl / BEE

Other
13 stars 3 forks source link

Cancel workflows leaves an orphaned neo4j instance #730

Closed pagrubel closed 6 months ago

pagrubel commented 9 months ago

The neo4j pid should be in the wf manager database. The neo4j instance should be killed if a workflow is cancelled.

Also, should the workflow be left in the database with a CANCELLED status?

pagrubel commented 9 months ago

If some tasks finished, the user may want the workflow left in the database.

pagrubel commented 9 months ago

Also, as we fix this, evaluate if we should cancel workflows from beeflow core stop and/or beeflow core reset

jtronge commented 8 months ago

A related problem is when someone submits a workflow but with the --no-start option. This will also leave an orphaned neo4j process, assuming that they don't eventually call beeflow start on the workflow.

pagrubel commented 7 months ago

This issue needs to be closed. We need to check everything above and see if we need to open other issues.

pagrubel commented 6 months ago

A related problem is when someone submits a workflow but with the --no-start option. This will also leave an orphaned neo4j process, assuming that they don't eventually call beeflow start on the workflow.

This does leave a neo4j instance when beeflow is stopped using beeflow core stop, but it isn't actually orpaned because after a beeflow core start the workflow can be started and it works (it can also be cancelled).

pagrubel commented 6 months ago

I'm closing this issue since PR #738 addresses this issue with orphaned neo4j instances. The related issue for beeflow core reset will be opened as a separate issue.