7.14.3.13 — September 2026

Phases

• Phase 7.4 complete — the replacement drives were validated as a set and are carrying production data.
• Phase 7.10 complete — the new platform now matches everything the old one did, plus tested restores and single sign-on.
• Phase 7.12 begins — configuration management is being rebuilt, so every machine's settings come from code rather than from hand edits.
• Phase 7.13 complete — the last hardware faults are cleared: write caching healthy on all three machines and firmware current.
• Phase 7.14 complete — both clusters are running at full strength on the new platform, spread across the machines as designed.

Services

• Released and documented Off-site backups of the platform's own data, encrypted end to end, sent to a remote site over a private tunnel and proven by a full restore test.
• Released Single sign-on console access for the hypervisor and backup hosts, with per-role permissions replacing the shared local logins.
• Added A feedback widget on the portal and the wiki, so a bug or a request can be filed without leaving the page you are on.

Stability

• Fixed The write-cache module and a failed fan on the third machine, ending the controller resets that had stalled a storage rebuild.
• Fixed Per-array write caching left switched off after that module replacement, on the array holding the cluster database, while the controller still reported healthy.
• Fixed The local backup target's long run of failures, traced to a drive reporting uncorrectable sectors rather than to the backup software.
• Fixed The feedback portal losing its colour scheme and falling back to the stock palette whenever the theme was saved.
• Upgraded Storage-controller and drive firmware to current across all three machines.
• Fixed Signing in to the management console granting no permissions at all, caused by a group-name mismatch that reported no error anywhere.
• Fixed Reverse name lookups returning nothing network-wide, so logs and monitoring show machine names again instead of bare addresses.
• Retrofitted All five switches so the isolated cluster network is no longer carried on ordinary device ports.
• Fixed A network profile that silently ignored its own settings and dropped the isolated cluster network, briefly costing one machine its primary cluster link.
• Added Reconstruct-write on the storage array, so saving a file no longer has to read the other disks first.
• Fixed Changelog updates reaching nobody since August: they were being marked as sent without anyone actually being notified.
• Fixed The management cluster having two of its three coordinators on one machine, so losing that machine no longer costs it quorum.

Documentation

• Documentation of the three-host firmware maintenance window, including the corrections found while running it.

Tracks in this release: Phases · Services · Stability · Documentation