jgrusewski
|
3bdefe86d5
|
feat(monitoring): production dashboards, remove Cockpit, add Kapsule logs
- Remove Cockpit remote_write from GitLab Prometheus (all metrics local)
- Uninstall k8s-monitoring Helm (Alloy no longer pushes to Cockpit)
- Deploy standalone node-exporter DaemonSet + kube-state-metrics
- Add 10 extra Prometheus scrape configs: node-exporter, kube-state-metrics,
kubelet, cadvisor, DCGM GPU, foxhunt services, Loki, Tempo, Promtail,
prometheus.io annotated pods
- Provision 9 Grafana dashboards via sidecar (ConfigMap-based):
overview, services, GPU/training, infrastructure, logs, CI/CD pipelines,
cluster overview, GPU overview, GitLab services
- Expand Promtail to ingest kube-system + GitLab runner + node-level logs
- Add dashboard.fxhnt.ai DNS record + nginx routing to Grafana
- Update CI deploy job to apply new monitoring manifests
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
|
2026-02-28 22:20:32 +01:00 |
|