ops-deploy

Ops: Deploy

Deploy the Rails application to remote environments.

Argument: $ARGUMENTS — environment (staging, production) and optional method (kamal, ci; default: kamal)

Safety

CRITICAL: Production deployments require explicit human confirmation before proceeding. Always ask for confirmation when $ARGUMENTS contains production.

Discovery

1. Read config/deploy.yml to discover the Kamal configuration: service name, registry, image, servers, accessories
2. Read config/deploy.staging.yml and config/deploy.production.yml for environment-specific overrides
3. Verify the specific environment-variable keys needed for the deploy are present in .env.staging / .env.production (or in SSM) without printing secret values
4. Read Dockerfile to understand the Docker build stages

CI/CD Path (Preferred)

The standard deployment path is via CI/CD — pushing to environment branches triggers auto-deploy.

# Deploy to staging via CI/CD
git push origin HEAD:staging

# Deploy to production via CI/CD (requires human confirmation first)
git push origin HEAD:production

Monitor the deployment via GitHub Actions:

gh run list --branch {environment} --limit 3
gh run watch {run-id}

Kamal Deployment (Manual)

Pre-Deploy Checks

1. Verify Kamal is installed:

   kamal version
   

2. Verify Docker builds locally:

   docker build -t {app_name}:test .
   

3. Check current deployment state:

   kamal details -d {environment}
   

4. Check for deploy lock:

   kamal lock status -d {environment}
   

   If locked from a previous interrupted deploy: kamal lock release -d {environment}

Deploy to Staging

kamal deploy -d staging

Deploy to Production

Requires explicit human confirmation.

kamal deploy -d production

Deploy with Specific Git Ref

kamal deploy -d {environment} --version {git-sha-or-tag}

Rollback

If a deploy causes issues, roll back to the previous version:

# List available versions
kamal app containers -d {environment}

# Rollback to previous version
kamal rollback {previous-version} -d {environment}

Post-Deploy Verification

After any deployment:

1. Health check the deployed environment:

   curl -sf -o /dev/null -w "HTTP %{http_code} in %{time_total}s" https://{app_host}/up
   

2. Verify ECS service stability (running count matches desired count):

   aws ecs describe-services \
     --cluster {cluster-name} \
     --services {service-name} \
     --region {aws-region} \
     --query 'services[0].{Running:runningCount,Desired:desiredCount,Status:status}' \
     --output table
   

3. Check for migration status (if migrations were included):

   kamal app exec --roles=web "bin/rails db:migrate:status" -d {environment}
   

4. Check logs for errors in the first 5 minutes (use ops-check-logs skill)

5. Verify Solid Queue workers are running (use ops-verify-jobs skill)

6. Verify OpenTelemetry traces are being exported (use ops-verify-telemetry skill)

Kamal Utility Commands

Command	Purpose
kamal details -d {env}	Show current deployment details
kamal app logs -d {env}	Tail application logs
kamal app exec --roles=web "bin/rails console" -d {env}	Open Rails console on remote
kamal audit -d {env}	Show deploy audit log
kamal env push -d {env}	Push updated environment variables
kamal lock status -d {env}	Check deploy lock status
kamal lock release -d {env}	Release a stale deploy lock
kamal traefik reboot -d {env}	Restart the Traefik proxy

Output Format

Report deployment result as a table:

Target	Environment	Method	Status	Verification
Rails app	staging	Kamal	SUCCESS/FAIL	/up returns 200
ECS tasks	staging	N/A	STABLE/UNSTABLE	running == desired
Solid Queue	staging	N/A	RUNNING/DOWN	workers have heartbeat