/deploy — Local Service Restart & Verify

Safely restart local Powernode services and verify they're healthy. This is for LOCAL development restarts, not remote/Docker deployment.

Step 1: Pre-flight Check

pwd
git status --short | head -10

Note any uncommitted changes — warn if there are unstaged migrations.

Step 2: Restart Services

Based on scope (default: all):

backend

bash scripts/reload-backend.sh

Backend uses SIGUSR2 hot reload (~30ms). No downtime.

worker

This cell is module-composed: the unit is powernode-<moduleID>-sidekiq.service (a UUID), never powernode-worker@default — a guessed name fails systemctl silently. Discover it, then restart and verify the discovered unit (fall back to powernode-worker@default only if discovery finds nothing, for a plain installer-shape host):

UNIT=$(systemctl list-units 'powernode-*-sidekiq.service' --no-pager --no-legend --plain | awk '{print $1}' | head -1)
sudo systemctl restart "${UNIT:-powernode-worker@default}"
systemctl is-active "${UNIT:-powernode-worker@default}"

Worker drains jobs before stopping (~28s), so the health check waits for it (Step 3).

worker-web (always restart with worker)

Same discovery, -worker-web instead of -sidekiq:

UNIT=$(systemctl list-units 'powernode-*-worker-web.service' --no-pager --no-legend --plain | awk '{print $1}' | head -1)
sudo systemctl restart "${UNIT:-powernode-worker-web@default}"
systemctl is-active "${UNIT:-powernode-worker-web@default}"

frontend

Same discovery, -frontend:

UNIT=$(systemctl list-units 'powernode-*-frontend.service' --no-pager --no-legend --plain | awk '{print $1}' | head -1)
sudo systemctl restart "${UNIT:-powernode-frontend@default}"
systemctl is-active "${UNIT:-powernode-frontend@default}"

all (default)

Restart in this order: backend (reload), worker + worker-web, frontend.

Step 3: Wait for Drain (worker only)

If worker was restarted, wait until its unit is active again before health checks (the ~30 s drain). A foreground sleep is blocked in this harness, so wait with the Monitor tool running a bounded until-loop on the discovered unit (shell state does not carry over, so discover it again):

UNIT=$(systemctl list-units 'powernode-*-sidekiq.service' --no-pager --no-legend --plain | awk '{print $1}' | head -1)
timeout 90 bash -c "until systemctl is-active --quiet '${UNIT:-powernode-worker@default}'; do sleep 2; done"

Step 4: Health Check

bash scripts/health-check.sh

If any checks fail, report which services are down and suggest remediation.

Step 5: Smoke Test (optional)

If all health checks pass, run a quick API smoke test:

curl -s http://localhost:3000/api/v1/health | python3 -m json.tool
curl -s http://localhost:4567/health | python3 -m json.tool

Step 6: MCP Connectivity (optional)

If MCP-related changes were made:

bash scripts/mcp-smoke-test.sh

Step 7: Report

| Service        | Status | Details          |
|----------------|--------|------------------|
| Backend        | ✅/❌  | reload/restart   |
| Worker         | ✅/❌  | drain time       |
| Worker Web     | ✅/❌  | port 4567        |
| Frontend       | ✅/❌  | port 5173        |
| API Health     | ✅/❌  | HTTP status      |
| Worker API     | ✅/❌  | HTTP status      |
| MCP            | ✅/❌/⏭ | if tested       |