Skip to content

KEDA Scaler Not Triggering

Symptom

A KEDA-managed workload stays at the same replica count even though queue depth, event lag, or the external trigger condition shows work pending.

Possible Causes

  • Workload Identity or secret-based authentication is failing.
  • The metric source is unreachable from the cluster.
  • The ScaledObject points at the wrong workload or has incorrect trigger metadata.
  • Another external metrics adapter is conflicting with the managed KEDA add-on.
  • The event source is healthy, but cooldown or min replica settings make the behavior look stuck.

Diagnosis Steps

flowchart TD
    A[KEDA scaler not triggering] --> B[Confirm KEDA add-on and operator pods are healthy]
    B --> C[Inspect ScaledObject status and events]
    C --> D{Auth or endpoint failure?}
    D -->|Yes| E[Fix Workload Identity or connectivity]
    D -->|No| F{Trigger metadata valid?}
    F -->|No| G[Correct trigger type target and values]
    F -->|Yes| H[Review cooldown min replicas and adapter conflicts]

  1. Verify the managed add-on is enabled.

    az aks show \
    --resource-group "$RG" \
    --name "$CLUSTER_NAME" \
    --query "workloadAutoScalerProfile.keda.enabled" \
    --output tsv
    Command Purpose
    az aks show Check whether the KEDA add-on is enabled.
    --resource-group Resource group that contains the AKS cluster.
    --name Name of the AKS cluster.
    --query Selects the KEDA enabled flag.
    --output Output format for the result.

  2. Confirm the operator and metrics API pods are healthy.

    kubectl get pods \
    --namespace kube-system \
    --selector app.kubernetes.io/part-of=keda-operator

  3. Inspect the ScaledObject and recent events.

    kubectl describe scaledobject.keda.sh <scaledobject-name> \
    --namespace <namespace>

  4. Look for typical failure patterns:

    • authentication or token errors,
    • connection failures to Service Bus, Storage, Event Hubs, Kafka, or Prometheus,
    • target workload name mismatch,
    • threshold values that never actually trip.

  5. If Workload Identity was enabled after KEDA, restart the operator so the expected environment variables are injected.

    kubectl rollout restart deployment keda-operator \
    --namespace kube-system

Resolution

  • Fix the federated identity credential, role assignment, or referenced secret path.
  • Correct the scaleTargetRef, trigger type, or trigger metadata.
  • Restore network reachability to the metric source.
  • Remove conflicting external metrics adapters and keep the KEDA add-on as the only external metrics server.
  • Increase observability by reviewing Kubernetes events alongside the ScaledObject status.

Prevention

  • Standardize scaler templates for each event source type.
  • Prefer Workload Identity over deprecated pod identity patterns.
  • Keep one owner for KEDA add-on lifecycle and scaler definitions.
  • Test trigger activation in pre-production with synthetic backlog before production rollout.

See Also

Sources