A workload is moved cleanly to a new namespace and only autoscaling stops responding while dashboards still show activity.
An HPA looks idle (metrics-adapter-cached-old-label-discovery)
Metrics exist, yet autoscaling remains flat because the adapter is still indexing the old label topology from before the move.
Scenario
What to check first
- Identify the primary failure signal in the Metrics Fresh, Adapter Discovery Cache Still Old 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
Check adapter discovery cache and metric shape before changing HPA thresholds.
Questions worth viewing together
Community-field Kubernetes problem inspired by adapter and HPA issue threads where cached discovery data survived a namespace move. Visible metrics do not prove the autoscaler is querying them through the same cached path.
Teams often blame the HPA object when the adapter is still presenting an old series map.
Autoscaling paths include adapter discovery state, not just raw metric availability.
Similar cases seen in the field