Rolling worker reboot (host OS)

Maand has no built-in reboot command. Use disable → stop → reboot → re-enable.

Related: disable-and-drain.md · rolling-deploy.md

Pattern A — disable worker, reboot, re-enable

# 1. Drain all jobs on the host
cat > workspace/disabled.json <<'EOF'
{ "workers": ["10.0.0.3"] }
EOF
maand build && maand deploy

# 2. Reboot (SSH from maand host)
maand run_command "sudo shutdown -r now" --workers 10.0.0.3

# 3. Wait for SSH (manual or scripted sleep)

# 4. Re-enable — remove 10.0.0.3 from disabled.json
maand build && maand deploy

# 5. Optional: verify all jobs after re-enable
maand health_check --jobs api,worker --wait

Disabled allocations keep deploy artifacts and KV; after reboot, deploy starts jobs again on re-enable.

Pattern B — rolling reboot across a fleet

maand run_command "sudo shutdown -r now" --workers 10.0.0.1,10.0.0.2,10.0.0.3 --concurrency 1

# Optional: verify all jobs after the rolling reboot
maand health_check --wait

No shell loop is required for the reboot command itself; maand run_command batches workers and honors --concurrency.

If you want strict disable -> deploy -> reboot -> re-enable -> deploy per worker, you still need per-worker disabled.json edits (scripted or manual).

Pattern C — parallel SSH only (no catalog drain)

For workers where a hard reboot without drain is acceptable:

maand run_command "sudo shutdown -r now" --workers 10.0.0.1,10.0.0.2 --concurrency 1

You can add --health_check to maand run_command to run job health checks after each command batch:

maand run_command "systemctl restart docker" --workers 10.0.0.1,10.0.0.2 --concurrency 1 --health_check

For reboot specifically, prefer explicit maand health_check --wait after the host returns, because health checks triggered immediately after shutdown -r now may run while the worker is still booting.

--concurrency 1 reboots one host at a time. After reboot, processes may be down until maand job start or maand deploy. Prefer Pattern A for production.

After reboot: sync check

maand job status api --allocations 10.0.0.3

If worker.json / update_seq mismatch, run maand deploy before maand job.