is there a way to programmatically terminate runs? I’m trying to figure out how to build a sensor that will terminate runs if they are taking longer than expected

There are two parts to this… If you have a run coordinator configured on your instance, you can call

cancel_run

on that run coordinator (if it supports that functionality)

cool, thanks - I’m using the

QueuedRunCoordinator

- I presume I can access that through

context.instance

?

You can do the latter without the former, but it might mean that there might be lingering computation chugging away even though Dagster has marked the run as canceled.

great, thank you!

I imagine it’d look something like this (untested):

if instance.run_coordinator.can_cancel_run(run_id):
    instance.run_coordinator.cancel_run(run_id)

# fetch the run to see if its state has changed since calling cancel
run = instance.get_run_by_id(run_id)
if not run.is_finished:
    # just mark run as canceled
    instance.report_run_canceled(run)

maybe also just worth noting the existence of the graphql api for this as well

ah yes, there’s a section in our graphql API docs for this exactly: https://docs.dagster.io/concepts/dagit/graphql#terminate-an-in-progress-run

@Dagster Bot discussion Programmatically terminate runs

Question in the thread has been surfaced to GitHub Discussions for future discoverability: https://github.com/dagster-io/dagster/discussions/8509