Need a more dynamic way to handle paused ScaledObjects during downtime and deployments

[cuixiaojun001]

Proposal

Feature Request

I propose adding an enhancement to KEDA to handle these scenarios more dynamically:

1. Dynamic Transition to Autoscaling: Allow KEDA to automatically transition from a paused state (autoscaling.keda.sh/paused) to normal autoscaling when the associated Deployment scales up from 0 replicas.

• Example: If the Deployment's replicas are increased from 0 (via manual scale, CI/CD tools, etc.), KEDA should detect this and automatically resume autoscaling without requiring manual annotation removal.
  Pause Behavior Configuration: Introduce a configuration option to specify whether the paused state should persist when replicas increase from 0. For example:

metadata:
  annotations:
    autoscaling.keda.sh/resume-on-scale: "true"

When set to true, KEDA would automatically resume autoscaling once the replica count is greater than 0.

Thank you for considering this request, and I look forward to your thoughts!

Use-Case

Background

In my current use case, I need to temporarily stop services by scaling the associated Deployment replicas to 0. During this downtime, I leverage KEDA's annotations to pause autoscaling:

metadata:
  annotations:
    autoscaling.keda.sh/paused: "true"
    autoscaling.keda.sh/paused-replicas: "0"

This works perfectly for ensuring that KEDA does not unintentionally scale up replicas during maintenance or downtime. However, when I need to resume the service or deploy a new version of the application, I face the following challenges:

• Manual intervention: Before the deployment, I must manually remove the paused annotation to allow KEDA to resume its autoscaling functionality. This adds significant complexity and risks missing this critical step.
• Frequent replica adjustments: In scenarios where the deployment needs to scale from 0 to any number of replicas (e.g., for different environments or scaling configurations), manually managing the paused annotation for each deployment becomes unmanageable.

Is this a feature you are interested in implementing yourself?

No

Anything else?

No response