rig-check

name

rig-check

description

System-domain health snapshot of the fuzz rig — host memory/disk/swap, systemd unit state, worker counts with uptime, OOM-kill events from dmesg, watchdog respawn rate, spurious kernel signals. Orthogonal to $check-in (which only covers fuzz-domain state). Use when the operator asks about VM/host health, OOM investigation, why workers keep dying, or invokes $rig-check. Equivalent of /rig-check in Claude Code.

Run the shared rig-check.sh. It auto-detects host:

• Mac / Linux host with orb → proxies commands through orb -m fuzzer
• Inside the fuzzing VM / host → runs commands directly

bash "$(git rev-parse --show-toplevel 2>/dev/null || echo .)/shared/rig-check.sh"

After the output prints, summarize anything flagged with [!]:

• Available memory < 512 MiB — under pressure; check afl-tmin or concurrent ASAN children.
• Disk free < 5 GB — archive findings/ or extend storage.
• Systemd service inactive — report which unit, suggest systemctl --user start ….
• OOM-kill count ≥ 5 — categorize victims:
  • total-vm ≈ 38 TB = ASAN child (unavoidable; ASAN shadow needs unbounded virtual). Expected noise.
  • total-vm few GB = fast-build child that escaped -m 1024 — check if afl-tmin is invoked with the right flag.
• Watchdog respawns > 10/hr — flap loop; drill into which role dies.
• Spurious kernel signals — non-fuzzer-child SIGSEGV / GPF; surface the listed process.

If nothing is flagged, report "rig healthy" — don't manufacture follow-ups.