Kubernetes HPA not scaling pods based on custom Prometheus metrics

The issue is that your Prometheus adapter configuration is using the wrong metric naming convention. The HPA expects metrics in the format 'pods/request_latency_p99' but the adapter is registering it as 'request_latency_p99_seconds'. Check your adapter ConfigMap - you need a seriesQuery that matches your actual Prometheus metric name and a metricsQuery that performs the correct aggregation. Here's a working config I use in production...