observability(runway): add artifact-registry storage and CDN dashboard

Adds a Storage and CDN detail dashboard for artifact-registry-gke, modeled on dashboards/registry/storage.dashboard.jsonnet. Four rows: GCS bucket (size by storage class, object count, billed byte-seconds, throughput by method), CDN cache (request rate by cache result, HIT/MISS share by response code), CDN bytes, and CDN latency (backend p50/p90/p99 by cache result, total latency, frontend TCP RTT).

The series come from the dedicated stackdriver-exporter-artifact-registry (argocd/apps!3344, !3376 (merged)) in the Mimir - Runway tenant. Selector notes for the reviewer:

  • project_id=~"gl-ar-storage-bucket-(prd|stg)" scoping is mandatory: the same metric families in this tenant also carry the Runway platform projects.
  • bucket_name=~"gitlab-.*-artifacts-registry": GCS attributes failed anonymous requests to the client-sent Host header, so unfiltered series include scanner hostnames.
  • These infra series carry environment = gprd/gstg (unlike the app metrics' production/staging); the dashboard uses a custom template accordingly.

Preview

Playground copy: https://dashboards.gitlab.net/d/jpereira-verify-argke-storage — all 26 panel queries verified against both environments via the datasource API. Cache-HIT panels are empty for now: CDN traffic to date is validation probes answering 403 MISS, so they populate with the first real cacheable traffic.

Scope note: this dashboard is the infrastructure vantage (Stackdriver GCS and CDN series). The registry sibling also carries app-vantage storage panels (per-action latency, retries, redirect ratios); AR has direct equivalents in its gitlab_artifact_registry_storage_* family (emitting in staging today), and those panels are a planned follow-up dashboard in this series rather than part of this MR.

Part of gitlab-org/ops/artifact-registry#354 (closed).

Edited by João Pereira

Merge request reports

Loading
Loading