Skip to content

v3.1.39

Choose a tag to compare

View all tags
@svaroqui svaroqui released this 06 Aug 23:39
· 113 commits to develop since this release
v3.1.39
4e3e5d3
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Highlights

Monitoring stability — repman no longer self-poisons on its own logs (#1666)

  • The dbjob now persists its stream offset — no more re-streaming the whole DB log every cycle (the ~100× collected-log inflation).
  • Fetched DB logs are tailed from EOF — no re-parsing the whole file on restart.
  • All repman-side logs always rotate + gzip + retain by size/count.
  • Bounded PFS snapshot retention — monitoring-pfs-snapshot-retention-days (default 2).

Reduced load on monitored databases

  • #1651 / #1652 — fixed a MariaDB Performance-Schema digest capture that was pathologically expensive and hammered monitored databases. Adds a master monitoring-performance-schema toggle, a configurable queries interval, and proper session isolation.
  • #1656 — the configurator now floors every allocated storage-engine buffer at 128M (fixes a case where a buffer pool could be provisioned at 0M and take the database down).

High availability / arbitrator

  • #1645 — arbitrator: multi-host support, bounded backend health/ping timeouts, and failover of backend DB connections.

Stability & resource leaks

  • #1657 — fixed a goroutine leak.
  • #1639 / #1640 / #1641 / #1643 — app health checks now refresh asynchronously; fixed leaked app-check HTTP connections; added per-app observability + staleness tracking; avoided marshal races.
  • #1647 — bounded DNS hostname resolution + configurable DNS timeout (avoids resolution hangs).
  • #1654 — fixed WARN0190/WARN0191 alert flapping (rejoin-catalog states added to pstates30).
  • #1646 — removed schema-monitoring debug logs that flooded the log.

Fixes

  • #1637 / #1638 — correct privilege matching for subnet-mask notation ('user'@'ip'/netmask).
  • #1649 — DB log-rotation cluster-setting toggles.
  • #1644 — dashboard proxy start route.
  • #1659 — self-heal a legacy .pull/<cluster>/plugins dir before the cloud18 pull.

Plugins / nightly

  • #1650 — nightly plugins published to the signer channel (fixes WARN0206 on nightly).

Known follow-ups (targeted for v3.1.40)

Log-rotation consolidation (#1671), db-log-rotate flag re-scope (#1667), event-log bounding (#1669), INFO state for un-managed DB-node logs (#1662), log-retention audit (#1668), query-sanitizer cache (#1664).

Assets 10
Loading