How do people prevent sensors from spiraling out of control and causing too many of same job to be running at the same time? Is there good way to set tags to prevent that?

Haven't used this with a sensor, however, we had a schedule get out of control. From the Dagster instance,

context.instance.get_run_records(RunsFilter(job_name="...", statuses=[...]))

hey @Daniel Mosesson you should be able to use the

run_key

parameter on

RunRequest

to ensure that two

RunRequests

with the same key only result in one run. https://docs.dagster.io/concepts/partitions-schedules-sensors/sensors#idempotence-and-cursors

But the issues is that I really want 10K runs, but I don't want more than 100 of them, to be running at any given time

oh i see, so i think what you want is to set concurrency limits https://docs.dagster.io/guides/limiting-concurrency-in-data-pipelines#limiting-concurrency-in-data-pipelines

for dagster in k8s open source, will this require a dagit restart?

yeah since you’d be making changes to

dagster.yaml

i believe you’ll need to restart dagster + the UI to see changes

Just for a update, I ended up going the "query the instance" route just because of the restart issue