This page lists all 18 Grafana dashboards that jakemon provisions, checked against the repository on 2026-10-11 (release v1.0.0). Dashboards are provisioned from jakemon/dashboards/, and the JSON in git is the source of truth. Change one by editing the JSON and running bin/deploy-central.sh. A change made in the Grafana UI is overwritten.
http://inference.lan:3000 on the LAN, or monitor.jakefear.com through Cloudflare Zero Trust. It needs a login.http://inference.lan:9090/api/v1/query, or copy its LogQL and run it against http://inference.lan:3100. Both are open on the LAN.uid is the dashboard's permanent address. URLs are /d/<uid>. Do not change a provisioned dashboard's uid in place: Grafana keeps serving the old one. The jakemon CLAUDE.md describes the rotation procedure.jakemon tag and a dropdown to every other dashboard.| uid | Title | Covers |
|---|---|---|
jakemon-home | jakemon — Overview | Grafana's default home page. Hosts and services up, alerts firing, plus tiles for the fleet, containers, logs, PostgreSQL, Wikantik, the edge and ollama. Each tile drills down to its dashboard. |
dashboards/infra/)| uid | Title | Covers |
|---|---|---|
jakemon-fleet | Fleet Overview | CPU, memory, load, network, disk I/O and fullest filesystem, per host |
jakemon-resources | Resource Consumption | Answers who is using each host. Includes: a fleet table of capacity versus use, with pressure stall information (PSI) columns; each host's CPU and memory split into containers, system services, user sessions, and kernel and other, which add up to the total; top containers and top systemd services for CPU, resident memory (RSS), disk and network; NICs, connections and PSI. |
jakemon-containers | Container Resources | Per Docker container name: CPU, RSS and working set, disk I/O, network, and memory against its limit |
jakemon-host-health | Host Health — silent failure modes | Failures that do not show as "down": failed systemd units, read-only filesystems, out-of-memory (OOM) kills; inode, file-descriptor and conntrack exhaustion; a disk-fill projection; thermals and temperature headroom, NIC errors, clock drift, reboots; disk SMART health. |
rYdddlPWk | Node Exporter Full | The community node-exporter dashboard, for raw host detail |
jakemon-postgres | PostgreSQL — Engine Internals | Internals of the PostgreSQL database that backs Wikantik on docker1 |
jakemon-logs | Logs | Container log search, error-log rate by container, and the top error-emitting containers |
jakemon-reliability | Reliability & Meta-monitoring | Includes: backup freshness (daily and off-site) and TLS days-to-expiry; container restart loops; every scrape target's health, including exporters that the ServiceDown alert excludes; Alertmanager delivery failures; the health of Prometheus and Loki themselves. |
dashboards/services/)| uid | Title | Covers |
|---|---|---|
wikantik-overview | Wikantik — Overview | Application metrics: JVM, HTTP status and latency, page activity, structural index, hybrid retrieval, agent and MCP tool traffic |
wikantik-host | Wikantik — Database & Search | PostgreSQL for Wikantik, embedding tables, query-embedder calls, latency and circuit-breaker state, and k6 load tests |
wikantik-traffic | Wikantik — Traffic (wikantik.com zone, human estimate, MCP) | wikantik.com zone traffic over the last 24 hours, from cf-analytics-exporter: countries, statuses, top URLs, and crawler and AI-crawler traffic with what each crawler was served. An estimated-human row (traffic split into human / bot / automation / suspect, and the wiki pages people read, ranked by distinct browsers). An MCP row from wiki-prod's own request counter: Claude Code burst timing and size per endpoint, status and latency, and per-tool frequency (LogQL over Wikantik's ECS event.action=mcp_tool_called lines, which Wikantik does not log yet) |
search-visibility | Search Visibility | Google, Bing and Yandex webmaster data (webmaster_*), URL Inspection coverage, SEO opportunities, and AI-referral log panels. See SearchVisibilityExporter. |
jakemon-cloudflared | Cloudflare Tunnel — edge | Tunnel liveness, HA connections, request throughput and errors, process resources, and the applied tunnel config revision |
edge-nginx | Edge nginx — cloudflare | Per-vhost traffic, from the edgemon JSON log of the static-site nginx on host cloudflare, plus stub_status internals and vhost probes |
craftercms-platform | CrafterCMS — Platform | The CrafterCMS engine, Studio, deployers and OpenSearch on docker2, from crafter-exporter |
craftercms-traffic | CrafterCMS — Traffic | Per-vhost traffic for jakefear.com and maiiavorobiova.com, from the delivery nginx's JSON access log |
jakemon-inference-gpu | Inference — RTX 4060 Ti + ollama | The GPU on inference, from DCGM: VRAM, thermals and utilization. Also ollama inference throughput and loaded models. Live again since 2026-10-03. |
/metrics scrape or health probe yet. That is the telemetry contract work for Roller.Alert rules live in central/prometheus/rules/. Each one names its runbook with runbook: Incident<AlertName>, and the runbooks are on this wiki. Alertmanager is loopback-only on inference (127.0.0.1:9093) and delivers to Telegram. See JakemonHub.