误删或者机器宕机，会导致 Etcd 数据的丢失或某个节点的 Etcd 数据异常时，请不要慌，认真看完此文，绝对有收获。当误删时，如何恢复数据，这个操作需求在实际环境当中是不可避免的。以下描述删除两个 namespace 下的 Pod，如何恢复对应 namespace 的数据。
备份etcd
ETCDCTL_API=3; etcdctl snapshot save snap.db --endpoints=https://172.16.230.84:2379 --cacert=/etc/kubernetes/ssl/ca.pem --cert=/etc/kubernetes/ssl/etcd.pem --key=/etc/kubernetes/ssl/etcd-key.pem {"level":"info","ts":1628844708.147097,"caller":"snapshot/v3_snapshot.go:119","msg":"created temporary db file","path":"snap.db.part"} {"level":"info","ts":"2021-08-13T16:51:48.169+0800","caller":"clientv3/maintenance.go:200","msg":"opened snapshot stream; downloading"} {"level":"info","ts":1628844708.1696196,"caller":"snapshot/v3_snapshot.go:127","msg":"fetching snapshot","endpoint":"https://172.16.230.84:2379"} {"level":"info","ts":"2021-08-13T16:51:48.441+0800","caller":"clientv3/maintenance.go:208","msg":"completed snapshot read; closing"} {"level":"info","ts":1628844708.467766,"caller":"snapshot/v3_snapshot.go:142","msg":"fetched snapshot","endpoint":"https://172.16.230.84:2379","size":"16 MB","took":0.320511222} {"level":"info","ts":1628844708.46799,"caller":"snapshot/v3_snapshot.go:152","msg":"saved","path":"snap.db"} Snapshot saved at snap.db
停止所有 master 上 kube-apiserver 服务
systemctl stop kube-apiserver
停止3台master上的 Etcd 运行
systemctl stop etcd
恢复备份
不同环境下，目录可能不一样，可以通过 systemctl status etcd 查看 Etcd 配置参数。特别需要注意 name、initial-cluster、initial-cluster-token、initial-advertise-peer-urls 和 data-dir 参数的值。
- 在第一台 Etcd 节点上，注意需要 ETCDCTL_API=3、name 值、IP 值、snapshot.db 文件目录和 data-dir 目录。
export ETCDCTL_API=3 一条指令，可以直接在终端上修改里面参数 etcdctl snapshot restore snapshot.db --name etcd1 --initial-cluster "etcd1=https://192.168.0.25:2380,etcd2=https://192.168.0.26:2380,etcd3=https://192.168.0.28:2380" --initial-cluster-token k8s_etcd --initial-advertise-peer-urls https://192.168.0.25:2380 --data-dir=/var/lib/etcd 和上面指令一样作用，把长的指令以换行形式展现 etcdctl snapshot restore snapshot.db --name etcd1 --initial-cluster "etcd1=https://192.168.0.25:2380,etcd2=https://192.168.0.26:2380,etcd3=https://192.168.0.28:2380" --initial-cluster-token k8s_etcd --initial-advertise-peer-urls https://192.168.0.25:2380 --data-dir=/var/lib/etcd 2021-01-19 11:17:06.773113 I | mvcc: restore compact to 96139 2021-01-19 11:17:06.800086 I | etcdserver/membership: added member 7370b1d3dc967c [https://192.168.0.25:2380] to cluster e4d7f96e88cc9d71 2021-01-19 11:17:06.800159 I | etcdserver/membership: added member 2ef3cfc4ca48ad38 [https://192.168.0.26:2380] to cluster e4d7f96e88cc9d71 2021-01-19 11:17:06.800190 I | etcdserver/membership: added member 3a0c86c4c744477c [https://192.168.0.28:2380] to cluster e4d7f96e88cc9d71
- 第二台和第三台 Etcd 恢复数据，同样需要改变 ETCDCTL_API=3、name 值、IP 值、snapshot.db 文件目录和 data-dir 目录。
export ETCDCTL_API=3 一条指令，可以直接在终端上修改里面参数 etcdctl snapshot restore snapshot.db --name etcd2 --initial-cluster "etcd1=https://192.168.0.25:2380,etcd2=https://192.168.0.26:2380,etcd3=https://192.168.0.28:2380" --initial-cluster-token k8s_etcd --initial-advertise-peer-urls https://192.168.0.26:2380 --data-dir=/var/lib/etcd 和上面指令一样作用，把长的指令以换行形式展现 etcdctl snapshot restore snapshot.db --name etcd2 --initial-cluster "etcd1=https://192.168.0.25:2380,etcd2=https://192.168.0.26:2380,etcd3=https://192.168.0.28:2380" --initial-cluster-token k8s_etcd --initial-advertise-peer-urls https://192.168.0.26:2380 --data-dir=/var/lib/etcd 2021-01-19 11:19:59.857363 I | mvcc: restore compact to 96139 2021-01-19 11:19:59.873793 I | etcdserver/membership: added member 7370b1d3dc967c [https://192.168.0.25:2380] to cluster e4d7f96e88cc9d71 2021-01-19 11:19:59.873837 I | etcdserver/membership: added member 2ef3cfc4ca48ad38 [https://192.168.0.26:2380] to cluster e4d7f96e88cc9d71 2021-01-19 11:19:59.873852 I | etcdserver/membership: added member 3a0c86c4c744477c [https://192.168.0.28:2380] to cluster e4d7f96e88cc9d71 export ETCDCTL_API=3 一条指令，可以直接在终端上修改里面参数 etcdctl snapshot restore snapshot.db --name etcd3 --initial-cluster "etcd1=https://192.168.0.25:2380,etcd2=https://192.168.0.26:2380,etcd3=https://192.168.0.28:2380" --initial-cluster-token k8s_etcd --initial-advertise-peer-urls https://192.168.0.28:2380 --data-dir=/var/lib/etcd 和上面指令一样作用，把长的指令以换行形式展现 etcdctl snapshot restore snapshot.db --name etcd3 --initial-cluster "etcd1=https://192.168.0.25:2380,etcd2=https://192.168.0.26:2380,etcd3=https://192.168.0.28:2380" --initial-cluster-token k8s_etcd --initial-advertise-peer-urls https://192.168.0.28:2380 --data-dir=/var/lib/etcd 2021-01-19 11:22:21.423215 I | mvcc: restore compact to 96139 2021-01-19 11:22:21.438319 I | etcdserver/membership: added member 7370b1d3dc967c [https://192.168.0.25:2380] to cluster e4d7f96e88cc9d71 2021-01-19 11:22:21.438357 I | etcdserver/membership: added member 2ef3cfc4ca48ad38 [https://192.168.0.26:2380] to cluster e4d7f96e88cc9d71 2021-01-19 11:22:21.438371 I | etcdserver/membership: added member 3a0c86c4c744477c [https://192.168.0.28:2380] to cluster e4d7f96e88cc9d71
- 三台 Etcd 启动
一条指令，可以直接在终端上修改里面参数 etcdctl --cacert=/etc/ssl/etcd/ssl/ca.pem --cert=/etc/ssl/etcd/ssl/node-node3.pem --key=/etc/ssl/etcd/ssl/node-node3-key.pem --endpoints=https://192.168.0.25:2379,https://192.168.0.26:2379,https://192.168.0.28:2379 endpoint health 和上面指令一样作用，把长的指令以换行形式展现 etcdctl --cacert=/etc/ssl/etcd/ssl/ca.pem --cert=/etc/ssl/etcd/ssl/node-node3.pem --key=/etc/ssl/etcd/ssl/node-node3-key.pem --endpoints=https://192.168.0.25:2379,https://192.168.0.26:2379,https://192.168.0.28:2379 endpoint health https://192.168.0.28:2379 is healthy: successfully committed proposal: took = 11.664519ms https://192.168.0.26:2379 is healthy: successfully committed proposal: took = 5.04665ms https://192.168.0.25:2379 is healthy: successfully committed proposal: took = 1.837265ms
总结
Kubernetes 集群备份主要是备份 Etcd 集群。而恢复时，主要考虑恢复整个顺序：
停止 Kube-apiserver--> 停止 Etcd--> 恢复数据 --> 启动 Etcd --> 启动 Kube-apiserver
注意：备份 Etcd 集群时，只需要备份一个 Etcd 就行，恢复时，拿同一份备份数据恢复。
参考：https://mp.weixin.qq.com/s/4b2COdr5q4SFfJTy3wl8gA