Google's container orchestration platform can now scale to massive clusters
Kubernetes 1.6, the latest version of Google's open source container orchestration platform, was released on Tuesday, featuring support for 5,000-node clusters. The 5,000-node cluster capability also will be added shortly to Google Container Engine, the company's Kubernetes-based automated container management system.
"For users who want to scale beyond 5,000 nodes or spread across multiple regions or clouds, federation lets you combine multiple Kubernetes clusters and address them through a single API endpoint," Kubernetes proponents said in a bulletin.
Support for 5,000 nodes is provided via the etcd 3.0 key value store, which is enabled by default in Kubernetes 1.6. Switching from etcd2 to etcd3 involves a full migration of data between different storage engines, with users having to stop the API from writing to etcd during the migration.
With version 1.6, integration with container runtimes is done through the Kubernetes CRI (Container Runtime Interface), for easier migration of runtimes via the kubelet node agent. Role-based access control, which defines secure default roles for control plane, node, and controller components, has graduated to a beta stage. The kubeadm cluster bootstrap tool, for installing Kubernetes on Linux, moves to a beta stage as well, as has the kubefed federation bootstrap tool, for cluster federation. "Docker remains the default runtime via Docker-CRI, which moves to beta," release notes state, and daemonsets, which ensure that nodes run a copy of a Kubernetes pod, can be updated via a rolling update in version 1.6.
With the 1.2 release a year ago, Kubernetes supported clusters of 1,000 nodes, while Kubernetes 1.3 scaled to 2,000-node clusters. Version 1.6 would follow by about three months the Kubernetes 1.5 release, which featured Windows Server and Windows Container support.
EmoticonEmoticon