The ‘Scrub-Type to Limitations’ Matrix
summary
A concrete matrix maps six scrub urgency levels to which limits each type honors; Squid implements many changes and Tentacles finalizes them. Key points: operator-initiated scrubs always run (except when PG state forbids), repairing scrubs ignore extended slip and CPU-load limits but respect per-OSD local concurrency, and OSD_max_scrubs queues instead of rejects so requests no longer fail. Deep-scrub timing uses a desired interval plus a configurable standard deviation (clamped to two stddev), and the default concurrency = 3 (was 1).
AI-written summary · about
chapters (9)
- 0:00Introduction to scrub matrix
- 4:40Scrub types and urgency levels
- 10:01Restrictions and configuration options
- 16:34OSD scheduling and recovery
- 21:16Changes in Tentacle scrubbing
- 24:12Shallow vs deep scrub updates
- 27:46Queue improvements in Tentacle
- 30:45Q&A on CPU load restrictions
- 34:21Future scrub improvements
related talks
39mKeynote: Lead Developers Town Hall
37mDefining and maintaining requirements in the Linux...
38mAgile, Lean & Rugged - The Paper Edition
20mNon-Linear Editing of Text-Based Screencasts
33mLow Context DevOps: Improving SRE Team Culture through Defaults, Documentation
33mPostgreSQL Observed—and Explained
07mTooling and Sharing Traceability Discussion
21mPrioritizing UX Content Work