Upgraded the server today. Nextcloud AIO from v13.1.0 to Hub 26 Spring, v33.0.5. All core containers healthy. Apache, Nextcloud, Database, Redis, Talk, Collabora, Imaginary, ClamAV, Whiteboard, Notify Push, Talk Recording. Twelve community containers deployed: Stalwart, Vaultwarden, Jellyfin, Plex, Fail2ban, Local AI, Notifications, SMB Server, Container Management, CalCardBackup, Borg Backup Viewer, and Collabora.
The big discovery was the RAID. Thirty terabytes of RAID 10, provisioned since December, never used. Six months of unused capacity. We configured Borg backup on it. First full backup took three hours. 567 GB source, 496 GB compressed. That's the safety net.
Fail2ban banned me from my own server during the upgrade. The Talk listener had been polling the Nextcloud API during the downtime. Failed requests piled up in the log. When Fail2ban came online with no whitelist, it read the log and banned my home IP. The server was reachable on the VPN and answered itself on localhost, but public access was blocked. The fix: whitelist admin IPs, service IPs, and partner IPs before enabling Fail2ban. The deputy needs to know who lives at the ranch before it starts arresting people at the front gate. That rule is now in the SOP.
We filed three strategic arcs. The Google ecosystem migration plan with a zero-loss guarantee. Servetus as an AIO community container with a five-phase roadmap. And the BigTech.sucks audit-to-liberation funnel.
We're eating our own cooking. Our email runs on our own iron, managed by our own AI. That's what we're selling.