Keynote: Ceph Developers AMA: Meet the Minds Behind Open Source Ceph
summary
mclock is now the default OSD scheduler but teams report rough edges, so defaults move from high client IO to balanced, recovery prioritization and device benchmarking are added and Crimson will add support; users report clusters up to 13,000 OSDs (practical scale a few thousand). Operators rely on upmap/PG remapper scripts to stage huge capacity changes while the balancer integrates that logic. Snapshot-based RBD mirroring is the focus (current scale ~100 images), and work continues on persistent client write cache, IPv6 telemetry, and slow‑op detection.
AI-written summary · about
chapters (21)
- 0:00Introduction and panel welcome
- 1:11mClock OSD scheduling status
- 5:29Introducing the Ceph team
- 7:38Scaling OSDs in production clusters
- 10:12Cluster health and upmap remapping
- 14:28Automatic storage tiering ideas
- 16:56Caching tiering and persistent caching
- 20:06RGW Keystone integration and roles
- 20:51Crimson and SeaStore energy efficiency
- 22:13Crimson and SeaStore hardware support
- 25:24AlmaLinux and Rocky Linux support
- 25:45Ceph consistency and stress reduction
- 27:09SEPH ADM and Rook tooling evolution
- 32:05DPU support for Crimson
- 34:34IPv6 support for Ceph Telemetry
- 36:00Ceph CSI dynamic provisioning
- 36:45RBD mirroring at scale
- 38:54RBD namespace quotas
- 40:17Package support in Ceph ADM
- 41:43Identifying unhealthy OSDs
- 45:18Handling denser storage media