Google introduces GKE Autopilot for hands-off Kubernetes

The new GKE Autopilot, generally available now, steps up the level of automation involved in Kubernetes management, down to eliminating all node management.

Google on Wednesday announced a new mode of operating Google Kubernetes Engine (GKE), its managed Kubernetes service. The new GKE Autopilot, generally available now, steps up the level of automation involved in Kubernetes management, down to eliminating all node management. 

Google's strength in containerized platforms, and notably Kubernetes, has helped the public cloud provider appeal to customers aiming to speed up their application delivery. According to a report from Flexera, GKE adoption among organizations reached 24 percent in 2020, up from 15 percent in 2019. 

Containers-as-a-service tools from Google's major public cloud competitors are also growing: AWS' Elastic Container Service and Elastic Kubernetes Service (ECS/EKS) experienced substantial adoption, with 54 percent using it, up from 44 percent in last year's Flexera cloud report. Another 24 percent planning to adopt ECS/EKS. Azure Container Service adoption reached 46 percent (up from 28 percent last year). 

GKE already offers a fully-managed control plane. With Autopilot mode, customers get a "hands-off fully managed Kubernetes experience," Google said in a blog post. 

In Autopilot, Google SREs manage nodes, including  provisioning, maintenance and lifecycle management. Autopilot nodes are locked down, preventing the kind of sysadmin-level modifications that make nodes unsupportable. The new mode gives you maintenance windows, as well as a pod disruption budget for maintenance flexibility. It comes with an SLA on pods in addition to the SLA on hosts and the control plane.

Autopilot also strives to provide a serverless experience, enabling customers to only pay for the pods they use. Customers are bill per second for vCPU, memory and disk resource requests. Autopilot provisions and scales compute infrastructure based on workload specifications and dynamic load. 

Cluster configurations in Autopilot are optimized for production. Additionally, the entire cluster infrastructure is secured, with GKE hardening guidelines and security best practices used to lock down individual nodes. 

Autopilot mode is initially compatible with DataDog's logging and monitoring tools, as well as GitLab's CI/CD tools. Google says there will be several more integrations in the coming months.

Prior and related coverage: