Public: anyone can read this; the crew can change it.
The data volume has under 15% free #13844
Check · Review · Land · Live done — Landed and live in production.
This pull makes the clone sweep remove an idle clone after three days instead of seven. That gives back some of the 22.8 GiB in crew, but it won't by itself bring free space back above 15%. What changed: CLONEIDLEMS in src/crew/orphans.ts goes from 7 to 3 days.
shipped production is serving it deploy 59903fa
This pull makes the clone sweep remove an idle clone after three days instead of seven. That gives back some of the 22.8 GiB in crew, but it won't by itself bring free space back above 15%.
- What changed:
CLONE_IDLE_MSinsrc/crew/orphans.tsgoes from 7 to 3 days.crewholds roughly 150 kept clones of about 150 MB each.node_modulesalready goes at one day, so what this removes is mostly sources and.git. - What stays safe: the sweep still keeps a clone that a live run holds. It also keeps one with uncommitted edits or unpushed commits. A run re-clones from the remote when the directory is gone.
- Docs and tests: the log line in
src/crew/dispatch.tsanddocs/features/disk-free.mdnow say three days. Intest/orphan-clones.test.tsI changed the clone ages to 4 days (stale or live) and 1 day (recent) so they still straddle the new line. - Checks run:
test/orphan-clones.test.tspasses (6 tests),bun x tsc --noEmitis clean, andbun scripts/gate.ts --fixfound nothing to fix. I did not run the full suite; the forge runs it on push. - What I chose not to do:
apps(33.7 GiB) andruns(8.0 GiB, already at its 8 GiB cap) were left alone. The three directories named in the issue add up to about 65 GiB, but 1.8 TiB is used. The rest is outside what my sandbox can read (/data/repos,backups,snapshotand others), so someone with access should size those.
To try it, push a clone idle for more than three days with nothing unpushed and watch the next hourly clone sweep log "removed clones idle for three days".
🤖 Generated with Claude Code
Approved. The constant, log line, doc and test ages (4d stale/live, 1d recent) all agree, and the stale-clone test would fail at the old 7-day window. It is a disk emergency fix, small and single-purpose, and it keeps the live-run and unpushed-work guards. No UI is touched, so there is nothing to see in the preview. Next two things: (1) rewrap the long line in docs/features/disk-free.md (the new three-days sentence runs well past the wrap used in the rest of the file) and move the 150-clones/22.8 GiB figures into the pull text only; (2) this does not by itself restore 15% free, so open a follow-up to size the unreadable directories (/data/repos, backups, snapshot) and cap apps/runs, since the 1.8 TiB used is mostly outside these three.
This pull makes the clone sweep remove an idle clone after three days instead of seven. That gives back some of the 22.8 GiB in
crew, but it won't by itself bring free space back above 15%.CLONE_IDLE_MSinsrc/crew/orphans.tsgoes from 7 to 3 days.crewholds roughly 150 kept clones of about 150 MB each.node_modulesalready goes at one day, so what this removes is mostly sources and.git.src/crew/dispatch.tsanddocs/features/disk-free.mdnow say three days. Intest/orphan-clones.test.tsI changed the clone ages to 4 days (stale or live) and 1 day (recent) so they still straddle the new line.test/orphan-clones.test.tspasses (6 tests),bun x tsc --noEmitis clean, andbun scripts/gate.ts --fixfound nothing to fix. I did not run the full suite; the forge runs it on push.apps(33.7 GiB) andruns(8.0 GiB, already at its 8 GiB cap) were left alone. The three directories named in the issue add up to about 65 GiB, but 1.8 TiB is used. The rest is outside what my sandbox can read (/data/repos,backups,snapshotand others), so someone with access should size those.To try it, push a clone idle for more than three days with nothing unpushed and watch the next hourly clone sweep log "removed clones idle for three days".
🤖 Generated with Claude Code
Closes #13843