From the front desk —

Now, the machine itself.

Everything else here is me telling you about the lab in plain language. This page is the lab telling you about itself — a sanitized, read-only snapshot pulled nightly from Prometheus. No addresses, no hostnames, no topology; just the shape of it.

sun@solar-hawk: ~/infra — btop live · sanitized
sun@solar-hawk:~$ infra --watch
51 / 57
VMs running
84 / 84
Scrape targets up
34 / 37
Uptime monitors up
74%
Cluster RAM free
~/cluster --load
mem█████████░░░░░░░░░░░░░░░░░░░░░░░░░26%
cpu███████████████████░░░░░░░░░░░░░░░55%
storage██████████████░░░░░░░░░░░░░░░░░░░░42%
targets up██████████████████████████████████100%

mem 393 / 1509 GB (1116 GB free) · storage 1.8 / 4.4 TB · cpu is mostly BOINC crunching donated compute, not production load.

~/observability
prometheus
30-day retention · 84 scrape targets · node, container & service metrics on every host.
loki
14-day logs · Promtail on every host · label-based filtering.
uptime-kuma
37 monitors · HTTP/TCP/DNS/cert checks · alerts → Gotify.
grafana
19 dashboards: compute, storage, containers, security, distributed compute.
● snapshot Jun 12, 2026, 11:30 PM (America/Detroit) · source: prometheus q to quit

How this stays safe

A scheduled job inside the lab reads the inventory and monitoring systems, runs every field through an allowlist (counts, hardware models, and statuses pass; everything else is redacted by default), and writes a clean JSON file the site rebuilds from. The public internet never touches anything inside the network — no addresses, no hostnames, no topology, no secrets.

Theme
Font