MariaDB migration recovery work needs a brief stable infrastructure integration window.
Repeated base changes delay the disposable cutover and recovery proof. That proof must pass before production migration can proceed.
Project progress
Total Infra
Highlights
MixMan PostgreSQL support and native checks are integrated. A fresh F34 backup restore passed.
Application support and source recovery safeguards advanced toward the coordinated PostgreSQL cutover. Production PostgreSQL acceptance and MariaDB retirement remain pending.
F34 passed disposable conversion and both crash recovery trials.
These results establish basic physical conversion and recovery evidence. Production still uses MariaDB. The complete production cutover owner, actual runtime handoff boundaries and live acceptance remain required.
MixMan staging now checks protected full qualification, and the recommendation branch has adopted lightweight development feedback.
Development checkpoints can stay inexpensive while complete tests, independent review and deployment-specific safety gates remain required.
F34’s migration repair passed full qualification and integrated. MixMan’s fresh PostgreSQL qualification also passed.
These checks remove known source defects and improve recovery evidence before a production cutover. Production still uses MariaDB.
The shared F34 recovery fix is active. A real conversion rehearsal found a legacy account compatibility gap.
The infrastructure blocker is resolved. The conversion now needs to preserve existing account values before recovery rehearsal can continue. Production still uses MariaDB.
Recent activity
MixMan PostgreSQL support and native checks are integrated. A fresh F34 backup restore passed.
Application support and source recovery safeguards advanced toward the coordinated PostgreSQL cutover. Production PostgreSQL acceptance and MariaDB retirement remain pending.
F34 passed disposable conversion and both crash recovery trials.
These results establish basic physical conversion and recovery evidence. Production still uses MariaDB. The complete production cutover owner, actual runtime handoff boundaries and live acceptance remain required.
MixMan staging now checks protected full qualification, and the recommendation branch has adopted lightweight development feedback.
Development checkpoints can stay inexpensive while complete tests, independent review and deployment-specific safety gates remain required.
F34’s migration repair passed full qualification and integrated. MixMan’s fresh PostgreSQL qualification also passed.
These checks remove known source defects and improve recovery evidence before a production cutover. Production still uses MariaDB.
The shared F34 recovery fix is active. A real conversion rehearsal found a legacy account compatibility gap.
The infrastructure blocker is resolved. The conversion now needs to preserve existing account values before recovery rehearsal can continue. Production still uses MariaDB.
MariaDB migration recovery work needs a brief stable infrastructure integration window.
Repeated base changes delay the disposable cutover and recovery proof. That proof must pass before production migration can proceed.
Deployed the heavy-work admission update and verified normal queued work under the new policy.
Allows admission under the approved headroom policy while keeping serialized heavy work and existing safeguards.
Six services in the coordinated UX update have passed delivery acceptance.
The accepted changes reduce lost work and make status, recovery, and keyboard interaction clearer. Live verification used read-only checks without production test records.
Restored all 14 hosted runners after reboot and identified two recurring boot failure paths.
The immediate runner outage is recovered. Reboot reliability still needs both coherent release selection and runtime preparation that does not depend on full application delivery.
The workspace audit batch has six merged and accepted fixes; Calendar and Home are verified live.
The reported bugs now have reviewed fixes, test evidence and closed lifecycle records.
Multiple Codex threads can now subscribe independently to the same workflow run.
The installed schema5 upgrade preserved129 receipts and52 batches. Each notification retains independent accounting, expiry and cancellation. Dashboard reader and agent guidance support the new distinction between waiting notifications and…
VPS callbacks now preserve waits through unreadable workflow observations.
Temporary observation failures retry within existing expiry instead of canceling the continuation. Installation preserved 126 receipts and 51 batches. Ordinary-chat preservation has live acceptance and issue505 is closed.
Installed explicit current-conversation enrollment for bounded Codex workflow notices.
Keeps completion reporting tied to authorized workflows while limiting automatic notices to one per run lifetime, three per workflow and five globally per rolling day, with one outstanding notice.
Bounded Codex workflow notices are installed and verified, with durable message limits and operator pause.
Software watches completion without model polling, while persistent one-per-run and rolling workflow/global limits prevent retry loops from creating unlimited new turns.
Codex workflow completion notices now have durable message limits and an explicit registration path.
Restored usable heavy-work queue visibility for agents and the operator dashboard.
Agents can follow current capacity and next-check guidance from application checkouts; operators no longer depend on an expensive full-workspace refresh for queue status. Shared capacity, FIFO, heartbeat, maintenance enforcement and alerti…
Wake publishing now supports readable project updates, optional supporting links, and explicit private posts. Delivery errors explain where a request stopped.
Useful progress can reach the project stream without becoming a mandatory completion checklist.
Integrated the Pyrophone-primary Zitadel cookie-backport build pipeline.
Heavy Login compilation moves off the VPS; exact-run artifact checks, packaged browser qualification and Login-only rollback-protected rollout retain the existing host lease and resource limits.
Published the isolated PostgreSQL qualification implementation and resolved its independent-review finding; broader verification is queued.
ProjectMan now has a reviewed source design for disposable PostgreSQL tests without sharing production database privileges. Nine focused contract tests pass; no runtime change or live PostgreSQL qualification is claimed.
Cleared premature planned-reboot maintenance advisory at the owner's request; scheduled reboot remains October 11 at 07:00 America/Chicago with no queue drain.
Verified maintenance is null and runner fleet active. The marker was advisory, not a lease admission lock; queue now reports host-pressure. No pressure thresholds, leases, runner configuration or reboot schedule changed.