Snapshots: Save & Restore

Warning: This page is only for ETCD version 3.x and higher!

For the original commands and more information on ETCD, see https://coreos.com/etcd/docs/latest/op-guide/recovery.html.

Take a snapshot

ETCDCTL_API=3 etcdctl \
    --endpoints $ETCD_ENDPOINT \
    snapshot save snapshot.db

(where snapshot.db is the name of the snapshot file to be created)

Restore a snapshot

Warning: Before restoring a snapshot, all ETCDs in the cluster must be stopped!

You must rename/remove the current data dir (probably /var/lib/etcd).

Be sure to provide all flags that are specified in, e.g., systemd unit file, Kubespray: /etc/etcd.env and others otherwise you may create issues for the ETCD cluster!

The command is looking about like that depending on what flags are used for your ETCD node:

# Run the command as `root` user after that use `chown` to correct ownership of files
ETCDCTL_API=3 etcdctl \
    snapshot restore snapshot.db \
    --name m1 \
    --initial-cluster m1=http://host1:2380,m2=http://host2:2380,m3=http://host3:2380 \
    --initial-cluster-token etcd-cluster-1 \
    --initial-advertise-peer-urls http://host1:2380
    --data-dir=/var/lib/etcd

# chown etcd:etcd -R /var/lib/etcd

This has to be done on all ETCD servers one by one with each having their own name given by flag as they were when the snapshot was taken.

Editing Kubernetes Objects

kubeadm