Use node_memory_wired_bytes (macOS-specific metric) instead of
node_uname_info to populate the instance dropdown. This prevents
Linux hosts like sifaka from appearing in the macOS dashboard.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
The metalmatze/transmission-exporter is unmaintained and has JSON parsing
issues with Transmission 4's API changes. Removing:
- Exporter sidecar from transmission deployment
- Transmission dashboard from Grafana
- Prometheus scrape config for transmission
- Metrics port from transmission service
TODO: Write custom transmission exporter
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add namespace dropdown variable for filtering all panels
- Add kube-state-metrics for k8s resource tracking
- Show pods, deployments, statefulsets counts
- Show memory/CPU requests by namespace
- Show namespace resource summary table
- Filter pod logs by selected namespace
- Add transmission-exporter sidecar for full metrics (speed, torrents, etc.)
- Add Prometheus scrape config for transmission metrics
- Update devpi dashboard to use blackbox probe metrics (status, response time, uptime)
- Restore transmission dashboard (will use exporter metrics)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Remove stale /opt/homebrew/var/loki from borgmatic backup (Loki migrated to k8s)
- Add Alloy k8s DaemonSet for automatic pod log collection to Loki
- Add blackbox probes for miniflux, kiwix, transmission, devpi, argocd
- Replace stale devpi/transmission dashboards with unified services health dashboard
- The new Alloy k8s deployment auto-discovers all pods including new ones
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
