Provision in advance/Pre heat

[yatanasov-hs]

I am using Keda 2.11 for scaling pods based on messages in AWS SQS. My Scaled obj definition looks like:

`apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
name: sample
namespace: sample-namespace
spec:
maxReplicaCount: 10
minReplicaCount: 0
scaleTargetRef:
name: sample
triggers:

• metadata:
  awsRegion: sample_region
  identityOwner: operator
  queueLength: "1"
  queueURL: sample_queue
  type: aws-sqs-queue
  `

Is there a way to always scale at calculated value + Constant. For example when there are no messages it will scale at Constant. When there are 2 messages it will scale to 2 + Constant. That would mean we will be always over provisioned by Constant pods. This will help me not have any delay when processing messages.

[JorTurFer]

Hello,
Does your workload process one message in parallel? I ask because I see queueLength and that means a pod per message. If this is your case, ScaledJobs currently support that feature. Using ScaledJob, you could set minReplicaCount:2 and you will have always 2 empty jobs, ready for processing the incoming mesages.
If you need to use ScaledObject for any reason, currently there isn't any built-in feature. This feature could help you in the near future, but it's not available yet.
If you are in rush, you could use Metrics API scaler to wrap SQS API with your own API, modifying the value with your custom logic. I mean, your API could just query SQS API and add 2 to the result.

[yatanasov-hs]

Correct, my worker processes 1 message at a time.
I'd rather not use jobs since there is caching in deployments which I would lose when running a job per message. Hoping to save time implementing custom api server.