Fleet/keep/middleware NATS_URL lists 10.10.10.21–23. Product streams
are replicas=3. px-worker worm/tree run on that cluster. Backup timer
and px-worker uptime watch are on. MOCK_VERAE and Zapier push unchanged.
NS1 is the Proxmox host. verae-proxmox creates LXC 510 (verae-px-worker
10.10.10.20 on vmbr1) with a private NATS proxy on 10.10.10.1:4222.
verae-uptime GET-watches public doors; verae-backup snapshots SQLite and
worm/tree data; verae-deploy does host-deps + checkout + npm ci.
Fleet overlays/ns1 are checked in (start.sh no longer rewrites JSON).
User systemd + linger for keep and fleet survive reboot.
Docs match current modules: public portal is access-web /portal/,
IAM and keep are listed, edge is loopback. Bootstrap clones every
Forgejo repo and installs deps per server type. Fleet starts IAM
and staff-session. NS1 all-in-one uses edge :13000 because :3000
is taken; archive workers stay with keep.
Restarts crashed units unless the admin console paused or stopped
them. Fleet writes per-instance intent.json. Tested on 138:
crash-restart, pause-hold, watch respawn of keep.
