Always be over provisioned based on message in a SQS queue

[yatanasov-hs]

We are scaling based on queue messages and have the requirement to always be over provisioned by 3 pods.
Our pod can process 3 messages at a time. I though I can use scalingModifiers but not clear what target means. E.g.:

apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
  name: scaledobject-foo
spec:
  advanced:
    scalingModifiers:
      formula: queuename + 3
      target: "3"
  maxReplicaCount: 10
  minReplicaCount: 0
  scaleTargetRef:
    name: foo
  triggers:
  - metadata:
      awsRegion: us-east-2
      identityOwner: operator
      queueLength: "3"
      queueURL: queue-url
    name: queuename
    type: aws-sqs-queue

This results in:
2 pods when there are 3 messages in the queue
1 pod when there are 0 messages in the queue.

Any help is well appreciated.

[zroubalik]

what about formula: queuename + 9 ?

[zroubalik]

The target (queueLength in case you don't use scaling modifier) is the key here, just set it to your threshold value, so 1 or 2.

[yatanasov-hs]

so target will equal my queueLength and the additional pods is queueLength * desired pods, i.e. 3 * 3 = 9, 3 * 2 = 6 ?

[zroubalik]

You don't need to specify queueLenth as it is ignored for scaling modifier, target is being used.
For the pods number calculation, please refer to https://keda.sh/docs/2.12/concepts/scaling-deployments/#triggers and check metricType section.

[yatanasov-hs]

thanks, makes sense now !

[zroubalik]

@gauron99 FYI, seems like there's a happy user of scaling modifiers 🚀