Prometheus Adapter¶

Ingests metrics from Prometheus via PromQL queries.

Configuration¶

prometheus:
  enabled: true
  url: http://localhost:9090
  scrape_interval_seconds: 30

Default PromQL Queries¶

Metric	Query	Entity Field
p95 Latency	`histogram_quantile(0.95, rate(http_request_duration_seconds_bucket[5m]))`	`latency_p95_ms`
p99 Latency	`histogram_quantile(0.99, rate(http_request_duration_seconds_bucket[5m]))`	`latency_p99_ms`
Throughput	`sum(rate(http_requests_total[5m])) by (service)`	`throughput`
Error Rate	`sum(rate(http_requests_total{status=~"5.."}[5m])) / sum(rate(http_requests_total[5m]))`	`error_rate`
Queue Depth	`sum(pending_requests) by (service)`	`queue_depth`

Entity Mapping¶

The Prometheus adapter creates ServiceEntity fragments with the following mapping:

job label       → entity_id
namespace label → owner_scope
cluster label   → scheduler_domain

Custom Metric Mappings¶

Override default queries in the config:

prometheus:
  enabled: true
  url: http://prometheus:9090
  metric_mappings:
    latency_p95: 'histogram_quantile(0.95, rate(my_custom_metric_bucket[5m]))'
    throughput: 'sum(rate(my_custom_rps[5m])) by (service)'