Cluster-scale Problem Solving
summary
Canonical Support Engineering uses Ceph to turn vague goals into measurable SLAs: they qualify and quantify workloads (example target: 100,000 IOPS), group by VM, measure backup chunk sizes (1 GB, 4 MB, 32 KB), run workload-specific benchmarks, and get formal signoffs to reach consensus. They document experiments, avoid the streetlight anti-pattern, reject outdated Nautilus configs in Quincy environments, and produce a short retrospective summary showing what was wanted, what was changed, and the verified results.
AI-written summary · about
chapters (8)
- 0:00Introduction to cluster problem solving
- 1:49Qualify and quantify cluster needs
- 3:18Benchmarking and workload analysis
- 5:27Achieving organizational consensus
- 6:12Technical problem solving process
- 7:36Documentation and retrospective
- 8:39Communication and active listening
- 9:43Q&A and closing remarks
related talks
40mJuggling Petabytes: Managing Ceph at Scale with Ceph-ansible
17mKeynote: Ceph Journey - A Perspective
44mContainer Management at Scale
28mA Welcoming Ceph Experience Improvements for First-Time Users or: How to Make Beginners Love Ceph
36m200 Clusters vs 1 Admin
29mTowards Sustainable Multi-Cluster Management in Real-World Academic Infrastructure
05mHealthier Ceph Clusters with Ceph-medic
20mManage Scaling, While Keeping Costs Under Control. Challenge Accepted.