Promotion & Rollback Playbook

Standard operating procedures for environment promotion and rollback on Lithosphere Makalu.

Promotion Flow

LOCAL  -->  TESTNET (Makalu)  -->  MAINNET
            auto on push           manual via promote.yaml
            deploy-simple.yaml     GitHub Environment approval

Staging is optional — promote directly from testnet to mainnet when staging is not configured.

Environment Promotion

Testnet (Automatic)

Testnet deploys automatically on every push to main:

1. CI passes (lint, build, test, gitleaks)

2. PR merged to main

3. deploy-simple.yaml triggers automatically

4. SSH deploy via bastion to EC2

5. Health checks verify API and Explorer

No manual intervention required.

Mainnet (Manual with Approval)

1. Identify the image tag to promote:

   # Find the tag from a successful testnet deploy
   # Check the deploy-simple.yaml run summary for the commit SHA
   # Or use the GHCR tag format: testnet-YYYYMMDD-HHMMSS

2. Trigger promotion via GitHub Actions:

   • Go to Actions > "Promote" workflow

   • Click "Run workflow"

   • Select:

     • Source: testnet

     • Target: mainnet

     • Image tag: the tag from step 1 (e.g. sha-abc1234 or testnet-20260307-143000)

   • Click "Run workflow"

3. Approval gate:

   • A reviewer listed in the mainnet GitHub Environment must approve

   • Configure reviewers: Settings > Environments > mainnet > Required reviewers

   • Reviewer verifies the image tag was tested on testnet

4. Deployment proceeds automatically after approval:

   • Images retagged with mainnet and mainnet-{timestamp}

   • SSH deploy to EC2 via bastion

   • Health checks run post-deploy

   • Automatic rollback on failure

Deployment Windows

Environment	Days	Hours (UTC)
Testnet	Any	Any
Mainnet	Tue-Thu	14:00-16:00

Freeze periods: Dec 20 - Jan 5, during chain events, 48h minimum between mainnet deploys.

Rollback Procedures

Automatic Rollback

Both deploy-simple.yaml and deploy.yaml include automatic rollback:

• A rollback snapshot is saved before each deploy (compose file, .env, image list)

• If health checks fail after deploy, the rollback job restores the snapshot

• The rollback job triggers automatically — no manual intervention needed

Manual Rollback (SSH)

If automatic rollback fails or you need to roll back outside of CI:

# 1. SSH to the server
ssh -o ProxyJump="ec2-user@44.218.142.100" ec2-user@10.0.10.16

# 2. Option A: Use the rollback script
cd /opt/lithosphere/Makalu
sudo bash scripts/rollback.sh

# 2. Option B: Manual restore from snapshot
cd /opt/lithosphere/Makalu
cp .rollback/docker-compose.yaml.bak docker-compose.yaml
cp .rollback/.env.bak .env
sudo docker compose down --timeout 30
sudo docker compose up -d --remove-orphans

# 3. Verify
curl -s http://localhost:4000/health
curl -s http://localhost:3100

Manual Rollback (Git-based)

Roll back to a specific commit:

ssh -o ProxyJump="ec2-user@44.218.142.100" ec2-user@10.0.10.16

cd /opt/lithosphere/Makalu
# View recent deploys
git log --oneline -10

# Checkout previous known-good commit
git checkout <commit-sha>
sudo docker compose up -d --build --remove-orphans

# Verify
curl -s http://localhost:4000/health

Mainnet Rollback Requirements

For mainnet rollbacks:

1. Notify the on-call engineer before starting

2. Execute the rollback (any method above)

3. Verify all health endpoints respond

4. File a post-incident review within 24h

Emergency Hotfix Fast-Path

For critical production issues:

1. Create a fix branch from main

2. Push and merge the fix PR (expedited review)

3. deploy-simple.yaml auto-deploys to testnet

4. Immediately trigger promote.yaml with:

   • skip_approval: true (emergency flag)

   • Image tag from the testnet deploy

Document the emergency in a post-incident review.

Blue/Green Deploy (Optional)

For zero-downtime deploys, use the blue/green script:

ssh -o ProxyJump="ec2-user@44.218.142.100" ec2-user@10.0.10.16
cd /opt/lithosphere/Makalu
sudo bash scripts/blue-green-deploy.sh

This:

1. Saves a rollback snapshot

2. Builds the new version as an isolated Docker Compose project

3. Health-checks the new version

4. Cuts over traffic only if healthy

5. Keeps the old version available for instant rollback

Health Check Endpoints

Endpoint	Port	Expected
API health	:4000/health	HTTP 200
API litho	:4000/api/litho	HTTP 200
GraphQL	:4000/graphql	HTTP 200 (POST)
Explorer	:3100	HTTP 200/302
Prometheus	:9090/metrics	HTTP 200

GitHub Environment Setup

To enable approval gates for mainnet:

1. Go to repo Settings > Environments

2. Create environment: mainnet

3. Add required reviewers (1-2 team leads)

4. Optionally add deployment branch rules (only main)

5. The promote.yaml workflow uses this environment for the approval gate

Similarly for staging if needed.

Troubleshooting

Deploy stuck / SSH timeout

# Check bastion connectivity
ssh ec2-user@44.218.142.100 echo "bastion ok"

# Check indexer via bastion
ssh -o ProxyJump="ec2-user@44.218.142.100" ec2-user@10.0.10.16 echo "indexer ok"

Containers not starting

ssh -o ProxyJump="ec2-user@44.218.142.100" ec2-user@10.0.10.16
cd /opt/lithosphere/Makalu
sudo docker compose ps -a
sudo docker compose logs --tail 50 api
sudo docker compose logs --tail 50 indexer

Disk full

sudo docker system prune -a --volumes -f
df -h /var/lib/docker