Hey, we have a job we’d like to run back to back -...
# ask-community
r
Hey, we have a job we’d like to run back to back - i.e when it completes, just fire up a new one (with the default run config). We currently use a schedule with a
QueuedRunCoordinator
but it can sometimes fill up the queue. Is Dagster Sensor the right tool for implementing that? Thanks
🤖 1
s
I thought the answer to this would be easy - simply conditionally yield a
RunRequest
as the last stage of a job. But I've just had a look at the documentation and it seems that `RunRequest`s can only be yielded by schedules and sensors. I haven't played around with this, but maybe you can yield a
RunRequest
from an op.
o
hi @Roei Jacobovich! You can't currently yield a RunRequest from an op (although that's a pretty interesting idea 🤔). A sensor is indeed the proper tool here. Here's an example answering a similar use case: https://github.com/dagster-io/dagster/discussions/7257 (here, they're launching runs in response to a job failing instead of a job successfully completing as in your case, but the structure is the same).
r
Hey @owen, @Stefan Adelbert, thanks for the response! I’ll check it out.
👍 1