ScaledJob "Target Average Value" prometheus

[rh-voice]

Hi there,

So I am trying to scale some jobs that run agents for our CI/CD; I can read pipeline demand metrics from prometheus (I have an exporter for that); Now this works fine with ScaledObjects; but I am trying to switch those to jobs.

My question is; what the "Target Average Value" attribute for prometheus ? because Threshold is not working; which is expected; it defines the threshold to start scaling for HPA not for spinning up jobs.

[JorTurFer]

Hey,
Sorry, I don't get you question, could you redo it?. The threshold is the target value per instance that the autoscaling tries to reach, if you set ie threshold: 5 and your prometheus returns 20, hypothetically you should have 4 instances, it's not something for starting from it. If you need that, we are introducing a new activation feature for next release, but current threshold is the amount of X per instance in jobs

[rh-voice]

Hmm, ok that makes sense. I am guessing its something in my polling interval; it might be polling faster than than the cooling down of that value.

could you elaborate on that new activation feature ? I've always wanted something webhook-like for scaledJobs; just an endpoint for the scaledJob to hit and request a job; that actually would solve my issue

[JorTurFer]

No no, the activation feature is not for that, it's just to not scaling from/to zero only if queue length is > 0, basically, because maybe you are not interested on having 1 instance only for 1 message and you could want to not scale from zero even you have any pending message.
This is the issue and this is the doc.
BTW, if you want to have and endpoint you can call to spawn jobs, you could use Metrics API and create the logic in an external REST API that KEDA calls using Metrics API scaler