Jetstream J3: requirements learned from each job's past runs issue #41 closed: completed andymccall opened this · 1 comment
andymccall set labels feature ·
andymccall referenced by change 8f211e20bea5 on main ·
andymccall
Done and deployed 2026-10-11 (runner d5fb3e43 on earth-1 and plex-1, hub 910fd849): finishes carry peak disk and memory, new jobs get need_disk from their last 5 runs (+25% +64 MiB), placement and the waiting reason use it. Test a_jobs_next_run_needs_the_disk_its_last_took.
andymccall closed ·
docs/jetstream.md §7 (refs issue #38). The runner's finish carries the job's peak disk (the blocks its scratch directory took: disks are sparse files there) and peak memory (the sandbox's VmHWM, sampled while it runs). The hub records them per job, gives each new job of the same repository and name a disk need of the most its recent runs used plus 25%, offers it only to a runner whose free disk (from its claim's facts, issue #40) covers that, and says so when it waits: 'needs ~12 GiB disk (its last runs used up to 10.4); plex-1 has 9 free'.