- commit
- 0ed7d0da406769e3683638cb4b0b0534fa817893
- parent
- 09eb5c46458f9ad9f98d5773ffcda99b63e179dd
- author
- dami <gazey55@gmail.com>
| M | PLAN.md | 17 | +++++++++++++++++ |
@@ -712,6 +712,23 @@ Also measured on the box while it generated: a play-through went
712712 revision 2 -> 245 in 60 s with two chunk processes at 140 MB peak each,713713 `/` answering in 0.13 s.714714 715+### Measured on the Fly box (2026-08-27) and the nightly's final shape716+717+On the shared-cpu-2x machine (gen-v6, band 24, jobs 2, GOMEMLIMIT718+256 MiB): ring 1 of the daily (+251 tiles) took **2,410 s** against719+~20 s on the Mac; the givens phase (SAT-heavy removal loop) dominates;720+max RSS per chunk 181 MB. Memory is a non-issue, CPU is: shared vCPUs721+are ~100x slower on this workload, so `--rings all` cannot fit a722+nightly window there. Decision: the app machine only serves (plus the723+small emergency field); the nightly is a **scheduled Fly machine with724+dedicated CPU** (`fly machine run --schedule daily`, performance-4x,725+dfw) that generates tomorrow's field with `--rings all` in minutes,726+POSTs it to `/push/<date>`, and stops — billed for minutes a day. The727+push route, built as a workaround, is the nightly's proper delivery728+path. Everything else from items 13-14 holds on the box: `/` with729+generating/missing states and yesterday's replay, territory-gated730+commits, idempotent join, restart-from-log, hosted tool list trimmed.731+715732 ### Status of the v2 items under this goal716733 717734 1 real packed run: done (v2 packings and v3 scatter fields played, logs