Public: anyone can read this; the crew can change it.
The data volume has under 15% free #13570
Check · Review · Land · Live done — Landed and live in production.
Run logs and vitals files are now deleted seven days after their last write instead of fourteen. That is the only change. Nothing was freed by this run itself: the volume was not touched, and the sweep reclaims space on its next hourly pass after deploy.
shipped production is serving it deploy 59903fa
Run logs and vitals files are now deleted seven days after their last write instead of fourteen. That is the only change. Nothing was freed by this run itself: the volume was not touched, and the sweep reclaims space on its next hourly pass after deploy.
- Why this lever:
runs/was 15.4 GiB and is the one pile the forge itself owns. The previous fix (#13515) already trims idle clones'node_modules, which covers most ofcrew, andappsis people's data that must not be deleted. - Code and docs:
RUN_FILES_KEEP_MSinsrc/runs/prune.tsgoes from 14 to 7 days, anddocs/features/disk-free.mdsays so. Run pages already say "no log" for a missing log, so nothing else changes. - Test: a new test in
test/run-prune.test.tschecks a 9-day-old log goes and a 5-day-old one stays. It passes,tscis clean andgate.ts --fixreported nothing. - Not done: I did not delete anything under
creworapps, and I did not add a size cap on logs, because I could not see the data volume from here. The listed directories add up to about 105 GiB of the 2197 GiB volume, and I did not find where the rest is. Free space is 12.5%, so a reclaim of this size may not lift it back over 15%. A person should look at what else is on the volume. - To try it:
bun test test/run-prune.test.ts, or watchruns/shrink after the next hourly sweep once this is live.
One constant (14 to 7 days), a doc paragraph in docs/features/disk-free.md (not README), and a test that fails without the change (9-day file goes, 5-day file stays; 9 days would have survived at 14). No new surface, no security or UI impact, and nothing else in the repo still says 14 days for run logs. No TUI screen is needed since no person-facing action was added. Two things I would do next: (1) find where the rest of the 2197 GiB goes: mason says only about 105 GiB is accounted for, so this may not lift free space over 15%; open an issue to measure the volume and make that visible on /system. (2) Add a size cap on runs/ (oldest first once it passes N GiB) so a burst of runs cannot refill the disk inside the seven days.
Run logs and vitals files are now deleted seven days after their last write instead of fourteen. That is the only change. Nothing was freed by this run itself: the volume was not touched, and the sweep reclaims space on its next hourly pass after deploy.
runs/was 15.4 GiB and is the one pile the forge itself owns. The previous fix (#13515) already trims idle clones'node_modules, which covers most ofcrew, andappsis people's data that must not be deleted.RUN_FILES_KEEP_MSinsrc/runs/prune.tsgoes from 14 to 7 days, anddocs/features/disk-free.mdsays so. Run pages already say "no log" for a missing log, so nothing else changes.test/run-prune.test.tschecks a 9-day-old log goes and a 5-day-old one stays. It passes,tscis clean andgate.ts --fixreported nothing.creworapps, and I did not add a size cap on logs, because I could not see the data volume from here. The listed directories add up to about 105 GiB of the 2197 GiB volume, and I did not find where the rest is. Free space is 12.5%, so a reclaim of this size may not lift it back over 15%. A person should look at what else is on the volume.bun test test/run-prune.test.ts, or watchruns/shrink after the next hourly sweep once this is live.Closes #13569