Troubleshooting

Intended audience: Anyone who is administering the Prompt Keda.

Known Issues

Issue	Description	Workaround

Monitoring

Redis Streams KEDA Performance Grafana Dashboard

Included in this dashboard are the following metrics. # keda_scaler_active Prometheus metric is set to 1 when the KEDA scalar is active and 0 when the scalar is not active. # keda_scaler_metrics_latency_second Prometheus metric shows the latency of the KEDA Scalar. A high latency suggests the scaler (Redis Streams for Prompt Processing) is # keda_scaled_job_errors_total The number of errors for each Scaled Job

Full metrics are here Update the path to the KEDA version running to see those metrics.

KEDA Cannot Connect to Kubernetes API

Symptoms: KEDA will report errors in the log that it cannot connect to the Kubernetes API.

Cause: This can happen if the Kubernetes API is unavailable or there is an issue with vCluster syncer.

Solution: This will sometimes correct itself. See Restarting KEDA Operator to restart the KEDA operator. If it does not help request for help in the usdf-infra-support Slack channel. Restarting DNS in the vCluster or restarting the vCluster syncer has resolved the issue in the past.