Skills
skills/epicweb-dev/epic-stack/epic-deployment

epic-deployment

SKILL.md

Epic Stack: Deployment

When to use this skill

Use this skill when you need to:

  • Configure deployment on Fly.io
  • Setup multi-region deployment
  • Configure CI/CD with GitHub Actions
  • Manage secrets in production
  • Configure healthchecks
  • Work with LiteFS and volumes
  • Local deployment with Docker

Patterns and conventions

Fly.io Configuration

Epic Stack uses Fly.io for hosting with configuration in fly.toml.

Basic configuration:

# fly.toml
app = "your-app-name"
primary_region = "sjc"
kill_signal = "SIGINT"
kill_timeout = 5

[build]
dockerfile = "/other/Dockerfile"
ignorefile = "/other/Dockerfile.dockerignore"

[mounts]
source = "data"
destination = "/data"

Primary Region

Configure primary region:

primary_region = "sjc" # Change according to your location

Important: The primary region must be the same for:

  • primary_region en fly.toml
  • Region del volume data
  • PRIMARY_REGION en variables de entorno

LiteFS Configuration

Configuration in other/litefs.yml:

fuse:
  dir: '${LITEFS_DIR}'

data:
  dir: '/data/litefs'

proxy:
  addr: ':${INTERNAL_PORT}'
  target: 'localhost:${PORT}'
  db: '${DATABASE_FILENAME}'

lease:
  type: 'consul'
  candidate: ${FLY_REGION == PRIMARY_REGION}
  promote: true
  advertise-url: 'http://${HOSTNAME}.vm.${FLY_APP_NAME}.internal:20202'
  consul:
    url: '${FLY_CONSUL_URL}'
    key: 'epic-stack-litefs_20250222/${FLY_APP_NAME}'

exec:
  - cmd: npx prisma migrate deploy
    if-candidate: true
  - cmd: sqlite3 $DATABASE_PATH "PRAGMA journal_mode = WAL;"
    if-candidate: true
  - cmd: sqlite3 $CACHE_DATABASE_PATH "PRAGMA journal_mode = WAL;"
    if-candidate: true
  - cmd: npx prisma generate --sql
  - cmd: npm start

Healthchecks

Configuration in fly.toml:

[[services.http_checks]]
interval = "10s"
grace_period = "5s"
method = "get"
path = "/resources/healthcheck"
protocol = "http"
timeout = "2s"
tls_skip_verify = false

Healthcheck implementation:

// app/routes/resources/healthcheck.tsx
export async function loader({ request }: Route.LoaderArgs) {
	const host = request.headers.get('X-Forwarded-Host') ?? request.headers.get('host')

	try {
		await Promise.all([
			prisma.user.count(), // Verify DB
			fetch(`${new URL(request.url).protocol}${host}`, {
				method: 'HEAD',
				headers: { 'X-Healthcheck': 'true' },
			}),
		])
		return new Response('OK')
	} catch (error) {
		console.log('healthcheck ❌', { error })
		return new Response('ERROR', { status: 500 })
	}
}

Environment Variables

Secrets in Fly.io:

# Generate secrets
fly secrets set SESSION_SECRET=$(openssl rand -hex 32) --app [YOUR_APP_NAME]
fly secrets set HONEYPOT_SECRET=$(openssl rand -hex 32) --app [YOUR_APP_NAME]

# List secrets
fly secrets list --app [YOUR_APP_NAME]

# Delete secret
fly secrets unset SECRET_NAME --app [YOUR_APP_NAME]

Common secrets:

  • SESSION_SECRET - Secret for signing session cookies
  • HONEYPOT_SECRET - Secret for honeypot fields
  • DATABASE_URL - Automatically configured by LiteFS
  • CACHE_DATABASE_PATH - Automatically configured
  • RESEND_API_KEY - For sending emails (optional)
  • TIGRIS_* - For image storage (automatic)
  • SENTRY_DSN - For error monitoring (optional)

Volumes

Create volume:

fly volumes create data --region sjc --size 1 --app [YOUR_APP_NAME]

List volumes:

fly volumes list --app [YOUR_APP_NAME]

Expand volume:

fly volumes extend <volume-id> --size 10 --app [YOUR_APP_NAME]

Multi-Region Deployment

Deploy to multiple regions:

# Deploy in primary region (more instances)
fly scale count 2 --region sjc --app [YOUR_APP_NAME]

# Deploy in secondary regions (read-only)
fly scale count 1 --region ams --app [YOUR_APP_NAME]
fly scale count 1 --region syd --app [YOUR_APP_NAME]

Verify instances:

fly status --app [YOUR_APP_NAME]
# The ROLE column will show "primary" or "replica"

Consul Setup

Attach Consul:

fly consul attach --app [YOUR_APP_NAME]

Consul manages:

  • Which instance is primary
  • Automatic failover
  • Data replication

GitHub Actions CI/CD

Basic workflow:

# .github/workflows/deploy.yml
name: Deploy

on:
  push:
    branches: [main, dev]

jobs:
  deploy:
    name: Deploy
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: superfly/flyctl-actions/setup-flyctl@master
      - run: flyctl deploy --remote-only
        env:
          FLY_API_TOKEN: ${{ secrets.FLY_API_TOKEN }}

Complete configuration:

  • Deploy to production from main branch
  • Deploy to staging from dev branch
  • Tests before deploy (optional)

Deployable Commits

Following Epic Web principles:

Deployable commits - Every commit to the main branch should be deployable. This means:

  • The code should be in a working state
  • Tests should pass
  • The application should build successfully
  • No "WIP" or "TODO" commits that break the build

Example - Deployable commit workflow:

# ✅ Good - Each commit is deployable
git commit -m "Add user profile page"
# This commit is complete, tested, and deployable

git commit -m "Fix login redirect bug"
# This commit fixes a bug and is deployable

# ❌ Avoid - Non-deployable commits
git commit -m "WIP: working on feature"
# This commit might not work, not deployable

git commit -m "Add feature (tests failing)"
# This commit breaks the build, not deployable

Benefits:

  • Easy rollback - any commit can be deployed
  • Continuous deployment - deploy any time
  • Clear history - each commit represents a working state
  • Faster recovery - can deploy any previous commit

Small and Short Lived Merge Requests

Following Epic Web principles:

Small and short lived merge requests - Keep PRs small and merge them quickly. Large PRs are hard to review, risky to merge, and slow down the team.

Guidelines:

  • Small PRs - Focus on one feature or fix per PR
  • Short-lived - Merge within a day or two, not weeks
  • Reviewable - PRs should be reviewable in 30 minutes or less
  • Independent - Each PR should be independently deployable

Example - Small, focused PR:

# ✅ Good - Small, focused PR
# PR: "Add email validation to signup form"
# - Only changes signup validation
# - Includes tests
# - Can be reviewed quickly
# - Can be merged and deployed independently

# ❌ Avoid - Large, complex PR
# PR: "Refactor authentication system and add 2FA and OAuth"
# - Too many changes at once
# - Hard to review
# - Risky to merge
# - Takes days to review

Benefits:

  • Faster reviews - easier to understand and review
  • Lower risk - smaller changes are less risky
  • Faster feedback - get feedback sooner
  • Easier rollback - smaller changes are easier to revert
  • Better collaboration - team can work in parallel on different small PRs

When PRs get too large:

  • Split into multiple smaller PRs
  • Use feature flags to merge incrementally
  • Break down into logical pieces

Tigris Object Storage

Create storage:

fly storage create --app [YOUR_APP_NAME]

This creates:

  • Tigris bucket
  • Automatic environment variables:
    • TIGRIS_ENDPOINT
    • TIGRIS_ACCESS_KEY_ID
    • TIGRIS_SECRET_ACCESS_KEY
    • TIGRIS_BUCKET_NAME

Database Migrations

Automatic migrations: Migrations are automatically applied on deploy via litefs.yml:

exec:
  - cmd: npx prisma migrate deploy
    if-candidate: true

Note: Only the primary instance runs migrations (if-candidate: true).

Database Backups

Create backup:

# SSH to instance
fly ssh console --app [YOUR_APP_NAME]

# Create backup
mkdir /backups
litefs export -name sqlite.db /backups/backup-$(date +%Y-%m-%d).db
exit

# Download backup
fly ssh sftp get /backups/backup-2024-01-01.db --app [YOUR_APP_NAME]

Restore backup:

# Upload backup
fly ssh sftp shell --app [YOUR_APP_NAME]
put backup-2024-01-01.db
# Ctrl+C to exit

# SSH and restore
fly ssh console --app [YOUR_APP_NAME]
litefs import -name sqlite.db /backup-2024-01-01.db
exit

Deployment Local

Deploy con Fly CLI:

fly deploy

Deploy con Docker:

# Build
docker build -t epic-stack . -f other/Dockerfile \
  --build-arg COMMIT_SHA=$(git rev-parse --short HEAD)

# Run
docker run -d \
  -p 8081:8081 \
  -e SESSION_SECRET='secret' \
  -e HONEYPOT_SECRET='secret' \
  -e FLY='false' \
  -v ~/litefs:/litefs \
  epic-stack

Zero-Downtime Deploys

Strategy:

  • Deploy to multiple instances
  • Automatic blue-green deployment
  • Healthchecks verify app is ready
  • Auto-rollback if healthcheck fails

Configuration:

[experimental]
auto_rollback = true

Monitoring

View logs:

fly logs --app [YOUR_APP_NAME]

View metrics:

fly dashboard --app [YOUR_APP_NAME]
# Or visit: https://fly.io/apps/[YOUR_APP_NAME]/monitoring

Sentry (opcional):

fly secrets set SENTRY_DSN=your-sentry-dsn --app [YOUR_APP_NAME]

Common examples

Example 1: Complete initial setup

# 1. Create apps
fly apps create my-app
fly apps create my-app-staging

# 2. Configure secrets
fly secrets set \
  SESSION_SECRET=$(openssl rand -hex 32) \
  HONEYPOT_SECRET=$(openssl rand -hex 32) \
  --app my-app

fly secrets set \
  SESSION_SECRET=$(openssl rand -hex 32) \
  HONEYPOT_SECRET=$(openssl rand -hex 32) \
  ALLOW_INDEXING=false \
  --app my-app-staging

# 3. Create volumes
fly volumes create data --region sjc --size 1 --app my-app
fly volumes create data --region sjc --size 1 --app my-app-staging

# 4. Attach Consul
fly consul attach --app my-app
fly consul attach --app my-app-staging

# 5. Create storage
fly storage create --app my-app
fly storage create --app my-app-staging

# 6. Deploy
fly deploy --app my-app

Example 2: Multi-region setup

# First region (primary) - 2 instances
fly scale count 2 --region sjc --app my-app

# Secondary regions - 1 instance each
fly scale count 1 --region ams --app my-app
fly scale count 1 --region syd --app my-app

# Verify
fly status --app my-app

Example 3: GitHub Actions workflow

# .github/workflows/deploy.yml
name: Deploy

on:
  push:
    branches: [main, dev]

jobs:
  deploy-production:
    if: github.ref == 'refs/heads/main'
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: superfly/flyctl-actions/setup-flyctl@master
      - run: flyctl deploy --remote-only --app my-app
        env:
          FLY_API_TOKEN: ${{ secrets.FLY_API_TOKEN }}

  deploy-staging:
    if: github.ref == 'refs/heads/dev'
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: superfly/flyctl-actions/setup-flyctl@master
      - run: flyctl deploy --remote-only --app my-app-staging
        env:
          FLY_API_TOKEN: ${{ secrets.FLY_API_TOKEN }}

Example 4: Deploy with migrations

# Create migration
npx prisma migrate dev --name add_field

# Commit and push
git add .
git commit -m "Add field"
git push origin main

# GitHub Actions automatically runs:
# 1. Build
# 2. Deploy
# 3. litefs.yml runs: npx prisma migrate deploy (only on primary)

Common mistakes to avoid

  • ❌ Non-deployable commits: Every commit to main should be deployable - no WIP or broken commits
  • ❌ Large, long-lived PRs: Keep PRs small and merge quickly - large PRs are hard to review and risky
  • ❌ Inconsistent primary region: Make sure primary_region in fly.toml matches the volume region
  • ❌ Secrets not configured: Configure all secrets before first deploy
  • ❌ Volume not created: Create the data volume before deploy
  • ❌ Consul not attached: Attach Consul before first deploy
  • ❌ Migrations on replicas: Only the primary instance should run migrations
  • ❌ Not using healthchecks: Healthchecks are critical for zero-downtime deploys
  • ❌ Deploy breaking changes without strategy: Use "widen then narrow" for migrations
  • ❌ Secrets in code: Never commit secrets, use fly secrets
  • ❌ Not making backups: Make regular database backups
  • ❌ FLY_API_TOKEN exposed: Never commit the token, only in GitHub Secrets

References

Preview Deployments (Inspired by Vercel Deploy Claimable)

Epic Stack can implement preview deployments similar to Vercel's deploy claimable pattern.

✅ Good - Preview deployments for pull requests:

# .github/workflows/preview-deploy.yml
name: Preview Deploy

on:
  pull_request:
    types: [opened, synchronize, reopened]

jobs:
  preview:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: superfly/flyctl-actions/setup-flyctl@master
      - name: Deploy preview
        run: |
          # Create or reuse preview app
          PREVIEW_APP="my-app-pr-${{ github.event.pull_request.number }}"
          flyctl apps list | grep "$PREVIEW_APP" || flyctl apps create "$PREVIEW_APP"
          
          # Deploy to preview app
          flyctl deploy --app "$PREVIEW_APP" --remote-only
        env:
          FLY_API_TOKEN: ${{ secrets.FLY_API_TOKEN }}
      - name: Comment preview URL
        uses: actions/github-script@v7
        with:
          script: |
            github.rest.issues.createComment({
              issue_number: context.issue.number,
              owner: context.repo.owner,
              repo: context.repo.repo,
              body: `🚀 Preview deployment: https://$PREVIEW_APP.fly.dev`
            })

✅ Good - Auto-cleanup preview deployments:

# .github/workflows/cleanup-preview.yml
name: Cleanup Preview

on:
  pull_request:
    types: [closed]

jobs:
  cleanup:
    runs-on: ubuntu-latest
    steps:
      - uses: superfly/flyctl-actions/setup-flyctl@master
      - name: Destroy preview app
        run: |
          PREVIEW_APP="my-app-pr-${{ github.event.pull_request.number }}"
          flyctl apps destroy "$PREVIEW_APP" --yes
        env:
          FLY_API_TOKEN: ${{ secrets.FLY_API_TOKEN }}

Environment Detection

✅ Good - Detect deployment environment:

// app/utils/env.server.ts
export function getDeploymentEnv(): 'production' | 'staging' | 'preview' | 'development' {
	if (process.env.NODE_ENV === 'development') {
		return 'development'
	}

	// Preview deployments
	if (process.env.FLY_APP_NAME?.includes('pr-')) {
		return 'preview'
	}

	// Staging environment
	if (process.env.FLY_APP_NAME?.includes('staging')) {
		return 'staging'
	}

	// Production
	return 'production'
}

✅ Good - Environment-specific configuration:

const env = getDeploymentEnv()

export const config = {
	production: env === 'production',
	staging: env === 'staging',
	preview: env === 'preview',
	development: env === 'development',
	
	// Preview deployments might have limited features
	features: {
		analytics: env === 'production',
		sentry: env !== 'development',
		indexing: env === 'production',
	},
}

Build Artifact Exclusion

✅ Good - Optimize Docker builds:

# other/Dockerfile
# Multi-stage build for smaller image size
FROM node:20-alpine AS base
WORKDIR /app

# Install dependencies
FROM base AS deps
COPY package*.json ./
RUN npm ci --only=production

# Build application
FROM base AS builder
COPY package*.json ./
RUN npm ci
COPY . .
RUN npm run build

# Production image
FROM base AS runner
ENV NODE_ENV=production

# Copy only what's needed
COPY --from=deps /app/node_modules ./node_modules
COPY --from=builder /app/build ./build
COPY --from=builder /app/public ./public
COPY --from=builder /app/server ./server
COPY --from=builder /app/other ./other
COPY --from=builder /app/prisma ./prisma
COPY --from=builder /app/package.json ./

# Exclude unnecessary files
# node_modules/.cache, .git, etc. are already excluded via .dockerignore

CMD ["npm", "start"]

✅ Good - Docker ignore file:

# .dockerignore (in other/)
node_modules
.git
.env
.env.*
!.env.example
*.log
.DS_Store
coverage
.vscode
.idea
*.swp
*.swo
*~
.cache
dist
build

Deployment Status and Monitoring

✅ Good - Deployment status tracking:

// app/routes/admin/deployment-status.tsx
export async function loader({ request }: Route.LoaderArgs) {
	const deploymentInfo = {
		appName: process.env.FLY_APP_NAME,
		region: process.env.FLY_REGION,
		environment: getDeploymentEnv(),
		commitSha: process.env.COMMIT_SHA,
		deployedAt: process.env.DEPLOYED_AT,
	}

	return { deploymentInfo }
}

Rollback Strategies

✅ Good - Quick rollback with Fly.io:

# List recent releases
fly releases list --app my-app

# Rollback to previous release
fly releases rollback --app my-app

✅ Good - Automated rollback on failure:

# fly.toml
[experimental]
  auto_rollback = true
  min_machines_running = 1
Weekly Installs
2
Repository
epicweb-dev/epic-stack
First Seen
2 days ago
Installed on
opencode2
codex2
claude-code2
antigravity2
gemini-cli2
windsurf1