Closed IbraAoad closed 1 month ago
Implementation is ready for review
TODO: Fix proxy image location + integration tests once https://github.com/canonical/metrics-k8s-proxy/pull/1 and https://github.com/canonical/metrics-proxy-rock/pull/1 are merged
Issue
The Istio ingress exposes metrics through a separate workload it deploys:
istio-ingress
: Accessible atistio-ingress-k8s-istio.{self.model.name}.svc:15020
, on port15020
at/stats/prometheus
.The main issue with integrating via the ingress's FQDN is that Kubernetes will enforce load balancing when the beacon is scaled, resulting in Prometheus receiving inconsistent metrics.
Solution
This PR adds metrics integration through the prometheus_scrape interface, utilizing the newly created metrics-k8s-proxy to aggregate these metrics into Prometheus via a unified endpoint.
Testing Instructions
up
metric for each theistio-ingress
pod.