TopolearnCloud Run is the quickest path from a container to a production URL, with no servers or clusters to manage.
Cloud Run runs stateless containers. You give it an image; it gives you an HTTPS URL, TLS, autoscaling (including to zero) and per-request or per-instance billing.
The contract for your container: listen on the port in the PORT environment variable (8080 by default), respond to HTTP, and keep no state on local disk you need later (the filesystem is in-memory and per instance).
Each deploy creates an immutable revision. Traffic can be split between revisions for canaries (gcloud run services update-traffic --to-revisions=...) and rolled back instantly.
Settings that matter:
- Concurrency: requests per instance (default 80). Lower it for CPU-heavy work.
- Min/max instances:
--min-instances=1removes cold starts;--max-instancescaps cost and protects your database. - CPU and memory, request timeout (default 5 minutes, up to 60).
- Ingress (all, internal, internal and load balancer) and authentication: by default callers need
roles/run.invoker;--allow-unauthenticatedmakes it public. - Service account the code runs as (use a dedicated one).
gcloud run deploy --source . builds your code with Cloud Build (Dockerfile or buildpacks) and stores the image in Artifact Registry for you. For run-to-completion work use Cloud Run jobs instead of services.
Example
# Dockerfile next to app.py: FROM python:3.13-slim ... CMD exec gunicorn -b :$PORT app:app
gcloud run deploy orders-api --source . --region=europe-west1 \
--service-account=orders-api@shop-dev-4821.iam.gserviceaccount.com \
--allow-unauthenticated --min-instances=0 --max-instances=20 --concurrency=40
# Service URL: https://orders-api-987654321098.europe-west1.run.app
# Canary: send 10% to the newest revision
gcloud run services update-traffic orders-api --region=europe-west1 \
--to-revisions=orders-api-00007-xyz=10,orders-api-00006-abc=90
gcloud run services logs read orders-api --region=europe-west1 --limit=20Remember
- Listen on $PORT and keep instances stateless
- Every deploy is a revision; split traffic for canaries
- Tune concurrency, min and max instances
- Services are private unless you allow unauthenticated access
Watch out for
- Doing background work after sending the response with request-based billing, where CPU is throttled outside requests.