Cloud Run
A fully managed, serverless platform for running containerized applications, scaling automatically to zero when idle and billing only for actual invocations. It combines the flexibility of running any container with zero OS management overhead, making it the right default for a standard stateless containerized web service.
Frequently Asked Questions
How is Cloud Run different from deploying to GKE?
Cloud Run removes cluster management entirely — there's no node pool to size, patch, or pay for while idle. You give it a container image and it handles scaling, including scaling to zero when there's no traffic, billing per-request rather than per-provisioned-capacity. GKE gives you more control (custom networking, DaemonSets, stateful workloads) at the cost of operating the cluster. For a stateless HTTP service with variable traffic, Cloud Run is usually less operational overhead.
What's a common mistake when moving a service to Cloud Run?
Assuming in-memory state or background threads survive between requests. Cloud Run can freeze or terminate an instance's CPU between requests and scale instances to zero, so any assumption of a persistent background job, in-memory cache warm across requests, or long-lived WebSocket beyond the request timeout will break intermittently. Also watch cold-start latency on scale-from-zero — set a minimum instance count if p99 latency on first request matters.