Autoscaler doesn't scale deployment when MinReplicas is set

[andrefrco]

Component(s)

collector

What happened?

Description

When using HPA with OpenTelemetryCollector, the deployment doesn't scale when HPA decides it needs more replicas. The issue is that GetInitialReplicas() always returns MinReplicas when autoscaler is configured, ignoring the Spec.Replicas field that HPA modifies.

opentelemetry-operator/internal/manifests/manifestutils/replicas.go

Lines 10 to 15 in 2584188

	func GetInitialReplicas(otelCol v1beta1.OpenTelemetryCollector) *int32 {
	if otelCol.Spec.Autoscaler != nil && otelCol.Spec.Autoscaler.MinReplicas != nil {
	return otelCol.Spec.Autoscaler.MinReplicas
	}
	return otelCol.Spec.Replicas
	}

Steps to Reproduce

1. Create an OpenTelemetryCollector with autoscaler configuration:

   apiVersion: opentelemetry.io/v1beta1
   kind: OpenTelemetryCollector
   metadata:
     name: backend-traces
     namespace: monitoring
   spec:
     autoscaler:
       minReplicas: 3
       maxReplicas: 6
       targetCPUUtilization: 80
       targetMemoryUtilization: 50
     # ... rest of collector config

2. Check HPA status after load:

    ❯ k get hpa backend-traces-collector
    NAME                       REFERENCE                               TARGETS            MINPODS   MAXPODS   REPLICAS   AGE
    backend-traces-collector   OpenTelemetryCollector/backend-traces   62%/50%, 42%/80%   3         6        5           386d

3. Check deployment replicas:

   ❯ k get deployment backend-traces-collector -o jsonpath='{.spec.replicas}'
   3⏎

4. Check OpenTelemetryCollector spec.replicas:

   ❯ k get otelcol backend-traces -o jsonpath='{.spec.replicas}'
   5⏎

5. Verify the issue:

   ❯ k get pods -l app.kubernetes.io/name=backend-traces-collector
   NAME                                        READY   STATUS    RESTARTS   AGE
   backend-traces-collector-59f8cbdbcb-grstd   1/1     Running   0          19h
   backend-traces-collector-59f8cbdbcb-k6268   1/1     Running   0          3h14m
   backend-traces-collector-59f8cbdbcb-rqvhg   1/1     Running   0          19h

Expected Result

When HPA decides it needs more replicas (e.g. 5), the deployment should scale to 5 pods, respecting HPA's decision.

Actual Result

Deployment stays at MinReplicas (3) even when HPA shows higher desired replicas. GetInitialReplicas() always returns MinReplicas when autoscaler is configured, preventing HPA from working properly.

Kubernetes Version

v1.29.4

Operator version

v0.135.0

Collector version

v0.135.0

Environment information

Environment

OS: Ubuntu 20.04.5 LTS (Kubernetes cluster nodes)

Log output

{"level":"DEBUG","timestamp":"2025-09-30T18:53:21Z","logger":"controllers.OpenTelemetryCollector","message":"desired has been updated","opentelemetrycollector":{"name":"backend-traces","namespace":"monitoring"},"object_name":"backend-traces-collector","object_kind":"&TypeMeta{Kind:,APIVersion:,}"}
{"level":"DEBUG","timestamp":"2025-09-30T18:53:21Z","logger":"events","message":"applied status changes","type":"Normal","object":{"kind":"OpenTelemetryCollector","namespace":"monitoring","name":"backend-traces","uid":"ee182001-1c2a-453b-aed4-097f47492ead","apiVersion":"opentelemetry.io/v1beta1","resourceVersion":"2062985166"},"reason":"Info"}

Additional context

This issue was previously discussed in #4255 (comment)

Tip

_{React with 👍 to help prioritize this issue. Please use comments to provide useful context, avoiding +1 or me too, to help us triage it. Learn more here.}

[gentoomaniac]

And I thought I would go crazy with this situation.
We experienced the same thing after an operator upgrade from 0.121.0 to 0.134.0

After some testing we could pin point it to 0.124.0 where it started breaking.

Was still trying to figure out where/why exactly.
Kudos for this bug report!