Running a Kubernetes API server from an etcd snapshot

Running a Kubernetes API server on a local workstation based on an etcd snapshot can be useful for debugging, and for understanding database backups. In a few simple steps this can be accomplished on Windows, Linux, or OSX.

First, locate the member/snap/db file, which is a backup of the etcd data directory. This file is usually stored in /var/etcd, /var/lib/etcd, or a backup server. Then, use this file to generate a new data-dir that can be run locally.

Prerequisites

• etcd and etcdctl for your workstation. Find a release here.
• kube-apiserver for your workstation. Go get k8s.io/kubernetes/cmd/kube-apiserver.

Setup

First, use etcdctl to create a copy of your data directory.

$ ETCDCTL_API=3 etcdctl snapshot restore kube-system-kube-etcd-0000/member/snap/db --name m1 --initial-cluster m1=http://localhost:2380 --initial-cluster-token etcd-cluster-1 --initial-advertise-peer-urls http://localhost:2380 --skip-hash-check

Now, run etcd against the generated data directory.

etcd --data-dir m1.etcd/

Finally, launch an API server that hits this etcd server running on localhost.

kube-apiserver --cert-dir ./kubernetes-temp --etcd-servers=http://localhost:2379 --service-cluster-ip-range 10.3.0.0/12

Now, curl http://localhost:8080/api/v1/events. Done!