Deploy FastAPI on Oracle Cloud's free tier

At a glance

Running FastAPI on Ampere ARM

uvloop and httptools both ship aarch64 wheels, so the standard extras install cleanly on Ampere. Long-lived WebSocket connections are the big win over serverless: nothing terminates the process between requests.

Steps

1. Expose a health endpoint. Cheaploy's blue/green deploy waits for this before shifting traffic.

   @app.get("/healthz")
   def healthz():
       return {"status": "ok"}

2. Add the Dockerfile. Two Uvicorn workers per OCPU is a good starting point for an async workload.
3. Deploy and watch the logs. Streamed logs and CPU/memory graphs are in the project dashboard.

Dockerfile for FastAPI on ARM64

FROM python:3.12-slim
WORKDIR /app
ENV PYTHONUNBUFFERED=1
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt "uvicorn[standard]"
COPY . .
EXPOSE 8000
CMD ["uvicorn", "app.main:app", "--host", "0.0.0.0", "--port", "8000", "--workers", "2"]

Need one tailored to your repository? The free Dockerfile generator writes an ARM-ready multi-stage build for you.

Common mistakes

• Blocking calls in async routes: A synchronous database driver inside an `async def` blocks the whole event loop. Use `def` routes or an async driver.
• No reverse proxy headers: Run with `--proxy-headers` so request.url and redirects use https, not http.
• Workers x memory: Each Uvicorn worker is a full process. Four workers of a 400 MB model will exhaust a 6 GB instance faster than you expect.

Related reading

• /guides/self-host-docker
• /compare/render-vs-cheaploy
• /oracle-cloud/always-free-arm
• All deployment guides