Hallo everybody, does anybody have an idea how to create and restore ETCD on HA . . .

Kubernetes-Slack Discussions
#1

Alex:
Hallo everybody, does anybody have an idea how to create and restore ETCD on HA cluster. Somehow it’s not working for me as expected. instance etcd-0 has the ip 10.10.10.100 and etcd1 10.10.10.101 , i created a backup using snapshot

etcdctl snapshot save snapshotetcd --endpoints=<https://10.10.10.100:2379> --cacert=/etc/kubernetes/pki/etcd/ca.crt --cert=/etc/kubernetes/pki/etcd/server.crt --key=/etc/kubernetes/pki/etcd/server.key

deleted a namespace foo and then restored the backup

1.moved the api-server manifests as mentioned in the documentation inorder to stop them
2.  etcdctl snapshot restore snapshotetcd --endpoints=<https://10.10.10.100:2379> --cacert=/etc/kubernetes/pki/etcd/ca.crt --cert=/etc/kubernetes/pki/etcd/server.crt --key=/etc/kubernetes/pki/etcd/server.key --data-dir /var/lib/etcd2
3. updated the  path: /var/lib/etcd2 on the master-0 where etcd-0 is runnung.
4. moved api-server manifest back
5. restarted kubelet
6. deleted other controll-plane components

everything is running back with /var/lib/etcd2 as directory in etcd-0, but the deleted namespace is not showing?
How can i be sure that etcd is using the /var/lib/etcd2 in etcd-0 and not /var/lib/etcd in etcd-1 ?

#2

unnivkn:
In an HA ETCD cluster, etcd’s should be in sync rt? Did you performed Step:3 on both etcd’s ?

#3

Alex:
@unnivkn i don’t thing this is required as etcdctl snapshot restore is applied to only one of them. If i perform step 3 on both of them, i need to copy the backup folder created from step 2 to the other master.

#4

unnivkn:
As per my understanding, in any High Availability Cluster, all the node’s should be in sync. Otherwise how it will act as HA cluster.
Also, please check inter etcd communication is enabled through port 2380 on both etcd’s on your etcd HA cluster.
eg: member1=http://10.0.0.2:2380,member2=http://10.0.0.3:2380