Release Journal
Changelogs
A transparent log of what shipped, why it shipped, and how each release improves production database reliability.
Platform
Manster scheduler rewrite
- Replaced cooperative workers with a deterministic queue model for stable p99 latency.
- Added per-node WAL pressure signals to prevent storage stalls during burst traffic.
- Improved hardware telemetry export for CPU cache misses and memory heat zones.
Runtime
Lightweight VM orchestration
- Provisioning now creates one lightweight VM per database to isolate noisy neighbors.
- Snapshot restore time dropped by 41% with the new block-map prefetcher.
- Database startup includes integrity checks before routing production traffic.
Network
Global ingress and failover
- Introduced region-aware ingress routing with health-weighted path selection.
- Added automatic read-only failover for replica-first recovery.
- Extended audit events for user actions, plan changes, and environment lifecycle.