Explicitly Supporting Stretch Clusters in Ceph
summary
Ceph gains a new connectivity mode for monitor leader elections that scores inter-monitor heartbeats so leaders reflect actual connectivity instead of lowest ID; this prevents continuous election flaps during netsplits. Election logic is split into a unit-tested election class so complex timing scenarios can be simulated quickly. OSD peering is extended to require replicas across data centers and OSDs are limited to local monitors, letting a two‑data‑center deployment continue serving I/O after one site loses connectivity in integration tests. A PR is pending to merge these changes.
AI-written summary · about
chapters (10)
- 0:00Intro
- 1:14RADOS WRITES
- 2:36STRETCH CLUSTER: DEFINITION
- 5:39LEADERS
- 6:20LEADER ELECTIONS
- 10:55MAKE CODE CHANGEABLE
- 12:41DEVELOP A NEW ALGORITHM
- 17:23OSD PEERING
- 18:19PEERING IN REAL LIFE
- 20:20DEAD DC SOLUTION: STRETCH MODE
related talks
32mStretch Clusters in Ceph: Algorithms, Use Cases, and Improvements
25mSupporting 3 Availability Zones Stretch Cluster
27mDetect, Decide, Fence: Resolving Netsplits in 3-AZ Stretch Cluster
34mModernizing Ceph Deployments at CERN: CephFS and Object Storage Across Data Centres
27mFrom Chaos To Clarity: Systematic Ceph Cluster Analysis for Rapid Troublesho...
49mIntro to Ceph, the Distributed Storage System
35mMaking Ceph Fast in the Face of Failure
12mCeph QoS: How to support QoS in distributed storage system