monitoring

how i know it all works. each tool does one job:

tool	job
gatus	checks that each service works, by what it returns
glances	host metrics, on every docker host and every proxmox node
dozzle	container logs from every host in one place
homepage	the dashboard that shows all of it
unpoller	UniFi device and client metrics, for prometheus
victorialogs	the log store: UniFi's events, by syslog

gatus replaced uptime kuma because its config lives in git, and uptime kuma's can't.

flowchart LR
    G[gatus<br>health checks] -->|results| H[homepage]
    L[glances<br>on every host] -->|metrics| H
    L -->|checked by| G
    A[dozzle agents<br>on every host] -->|logs| D[dozzle hub]
    D -->|tile| H

• homepage shows gatus's results on the plumbing tab, and each host's glances on the infrastructure tab
• gatus checks the other tools too: every glances, every dozzle agent and the dashboard itself
• all of it deploys from git except glances on the proxmox nodes, which i install by hand. gatus and homepage read their own config from git, so a merged change is live without a restart