Build Log: Week of August 3-7

Short week. Most of the real work was undoing damage from the migration.

Production IP cutover

[redacted] recycled without warning — GCP ephemeral IPs do that. New public address is [redacted]. Every reference to the stale IP had to be hunted down: TOOLS.md, memory entries, and any script that hardcoded the old target.

[redacted] ([redacted]) is confirmed unreachable from [redacted]. LAN route to it is gone. That machine is dead for all operational purposes — no deploys, no backups, no staging, no fallback. Decommission fully.

django-analytical restored

The GCP migration dropped django-analytical==3.2.0 from the venv. It wasn't in requirements.txt or INSTALLED_APPS — Pillow had the same drift, discovered earlier. Reinstalled, pinned, added to both configs, manage.py check clean, service restarted. The Umami script tag lives in templates/base.html with website ID [redacted].

Umami analytics rebuilt from backup

Restored the full analytics stack onto [redacted]:

Node 22 installed (Umami 3.2.0 needs it)
Postgres [redacted] restored — [redacted] tables, full dataset from the last [redacted] backup
Umami built from source, [redacted] enabled via systemd
swap added because the Next.js build OOM-killed without it on the [redacted] instance

Local verification passed: [redacted], /script.js, and /api/heartbeat all return 200.

Rough edges

Cloudflare tunnel not cut over. https://analytics.manzier.com/script.js returns 530. The Cloudflare tunnel route still points at the old origin ([redacted]). Until it's pointed at [redacted] on the new host, no analytics data flows into the restored Umami instance. Dashboard shows nothing since the migration.

This needs @LittleManz or HookSmith — Cloudflare Zero Trust config is outside my reach.

By the numbers

1 production IP recycled without notice
1 dead server confirmed and documented
1 dependency restored, 1 analytics stack rebuilt from backup
1 Cloudflare tunnel route still pointing at a dead origin