Changelog
Notable changes to planmycluster, newest first.
Subscribe to changes as a feed0.3.0
- Shared links now carry the plan after # instead of ?, so it never reaches a server. Links in the old format no longer open.
- A new start page that shows a calculated example result and explains what the planner answers.
- An about page and a page that states openly how the site is funded and what will not happen.
- The sitemap now lists only pages that search engines may index, and an llms.txt describes the site for AI assistants.
0.2.0
- Placement groups can be set by hand per pool, and the failure simulation shows how many of them a failure leaves untouched, degrades, blocks or loses.
- SAS SSDs can be planned as their own class, with the higher sustained throughput of the 12 Gbit/s interface in the recovery estimate. Ceph still sees them as ssd, and the generated commands say so.
- The cluster layout now shows real drive bays with status lights, each drive individually selectable, and a list of the result sections so the generated commands are one click away instead of a long scroll.
- Failure simulation for any number of nodes and OSDs at once, with the status recalculated from all of them together.
- Data loss is now reported as such: a cluster that can only place one copy loses everything when the host holding it goes down.
- Erasure coded pools are judged correctly after a failure: rebuilding needs k surviving chunks, so a pool below that is reported as lost instead of self-healing.
- A quieter motion layer across the site: cards and lists enter with a short animation, a figure that just changed lights up briefly, and all of it turns off when your system asks for reduced motion.
0.1.0
- Cluster capacity planning: raw, usable and planned (85 %) capacity for a mix of nodes, OSDs and pools.
- Placement group recommendations per pool, following the pgcalc formula and the power-of-two rounding rule.
- Redundancy findings for size/min_size and erasure coding (k/m), including failure domain and topology checks.
- Failure simulation for a single node or OSD, split into data availability and self-healing.
- A rough recovery time estimate, always shown as a range, never a single number.
- Generated pveceph and native ceph commands for the exact layout entered, with a "show all parameters" toggle.
- Example presets (homelab, SMB, HDD archive with erasure coding, and a deliberately dangerous setup) used throughout the learn pages.
- A shareable permalink that encodes the whole plan in the URL, with no server-side storage.
- Learn articles, glossary, FAQ and a rendered methodology page, in English and German.