oliver-kriska/deploy

Elixir/Phoenix Deployment Reference

Quick reference for deploying Elixir/Phoenix applications.

Iron Laws — Never Violate These

1. Config at runtime, not compile time — Secrets in config.exs get baked into the release binary. Use runtime.exs with env vars so secrets are resolved at boot
2. Graceful shutdown ≥ 60 seconds — Shorter timeouts kill in-flight requests and WebSocket connections mid-operation, causing data loss for users
3. Health checks required — Without startup/liveness/readiness endpoints, orchestrators can't distinguish a booting node from a dead one, leading to cascading restarts
4. SSL verification for database — Skipping verify: :verify_peer allows MITM attacks between your app and database; production data traverses the connection
5. No CPU limits — The BEAM scheduler assumes it owns all cores; cgroups CPU limits cause scheduler collapse where the VM thinks it has more cores than it can use, leading to latency spikes

Quick Configuration

runtime.exs (Essential)

if config_env() == :prod do
  database_url = System.get_env("DATABASE_URL") || raise "DATABASE_URL is required"
  secret_key_base = System.get_env("SECRET_KEY_BASE") || raise "SECRET_KEY_BASE is required"
  host = System.get_env("PHX_HOST") || raise "PHX_HOST is required"

  config :my_app, MyApp.Repo,
    url: database_url,
    pool_size: String.to_integer(System.get_env("POOL_SIZE") || "10"),
    ssl: true,
    ssl_opts: [verify: :verify_peer]

  config :my_app, MyAppWeb.Endpoint,
    url: [host: host, port: 443, scheme: "https"],
    http: [ip: {0, 0, 0, 0}, port: String.to_integer(System.get_env("PORT") || "4000")],
    secret_key_base: secret_key_base,
    server: true
end

Health Check Plug

def call(%{path_info: ["health", "readiness"]} = conn, _opts) do
  case Ecto.Adapters.SQL.query(MyApp.Repo, "SELECT 1", []) do
    {:ok, _} -> send_resp(conn, 200, ~s({"status":"ok"})) |> halt()
    {:error, _} -> send_resp(conn, 503, ~s({"status":"error"})) |> halt()
  end
end

Quick Decisions

Platform Choice

| Need | Use | |------|-----| | Simple, managed | Fly.io | | Enterprise, existing K8s | Kubernetes | | Custom infrastructure | Docker + your orchestrator |

Resource Limits

| Resource | Recommendation | |----------|----------------| | CPU | NO LIMITS (BEAM scheduler issues) | | Memory | Set limits (256Mi-512Mi typical) | | Graceful shutdown | ≥ 60 seconds |

Deployment Checklist

• [ ] All secrets from environment variables in runtime.exs
• [ ] server: true in endpoint config
• [ ] SSL verification for database connections
• [ ] Health endpoints: /health/startup, /health/liveness, /health/readiness
• [ ] Graceful shutdown period ≥ 60 seconds
• [ ] No CPU limits (memory limits only)
• [ ] Migrations in deploy process

Asset Pipeline Notes

Phoenix 1.8 uses esbuild + tailwind (no Node.js required):

• Config in config/config.exs under :esbuild and :tailwind
• mix assets.deploy builds for production
• mix assets.setup installs binaries on first run
• Custom JS bundlers: configure in config/config.exs

References

For detailed patterns, see:

• ${CLAUDE_SKILL_DIR}/references/docker-config.md - Multi-stage Dockerfile, best practices
• ${CLAUDE_SKILL_DIR}/references/flyio-config.md - fly.toml, clustering, commands