A platform team standardizes labels and later an HPA stops seeing metrics even though the backing monitoring query still returns data.
An HPA sees no metrics after a label cleanup because the metrics adapter cache still serves discovery data for the old label set while the new selector never becomes visible
The workload is healthy, yet scaling stops because the adapter did not refresh the metric discovery contract after a label schema change.
Scenario
What to check first
- Identify the primary failure signal in the The Metric Name Stayed the Same but Discovery Still Remembered the Old Labels scenario.
- Separate visible symptoms from the underlying technical dependency.
- Describe the safest recovery path and the follow-up prevention work.
Checking checklist
- Summarize the current impact and the last known change.
- Collect direct evidence from logs, runtime state, and configuration before changing anything.
- Separate immediate recovery from permanent prevention work.
Recovery and prevention
Compare adapter discovery output and selector assumptions before rewriting the HPA object.
Questions worth viewing together
Community-field Kubernetes problem inspired by scaling discussions where metrics adapter discovery stayed stale after a label cleanup. Autoscaling failures often come from stale metric discovery metadata, not from the metric source itself.
Teams often blame Prometheus when the adapter cache is still advertising the old label contract.
Metrics adapters need explicit refresh validation after relabeling efforts.
Similar cases seen in the field