Notes
The backup cycle closes: the full circuit runs in production
Javier unblocked the headquarters IP in Cloudflare and the whole chain
1 min read
Monday brought the resolution of Friday's issues. Javier unblocked the headquarters IP in Cloudflare — that was what was blocking port 443 to the catalog — and the whole chain was verified over SSH on both servers.
The full circuit, for the first time in production
The promise was this: a new release comes out that requires a new updater, the updater updates itself on its own, and the core button leaves the features ready. Monday was its real debut. The catalog published 0.1.41 and both servers applied it on their own; the update-available field went to false on both.
The best part was seeing the chained auto-update: both servers moved from 0.1.26 to 0.1.28 on their own, in two hops, with the "updating automatically" notice and the unit restart included. Without a single SSH command. Remembering the three bugs we caught on Friday — the protocol, the private /tmp, and the cgroup — seeing the full mechanism running in production has a different flavor.
Backups operational on both servers
With the path clear, backups became operational: Be Coach with three copies — the latest from that morning — and Web Base with fifteen, about 8 MB per copy. The restore button and deletion are active. The cycle that was promised is closed and verified: new release, updater that updates itself, core button, features ready.