Restore ECTD from a backup

rahulshettigar13 · July 9, 2024, 11:02am

What is the correct way to restore ETCD? the practise video has a different appoard, the solution after the practise video has a different approach as well

github.com

kodekloudhub/community-faq/blob/main/docs/etcd-faq.md#how-do-i-restore-a-backup

# ETCD FAQ

**IMPORTANT** Following recent student comments about the exam backup/restore questions,  this FAQ is currently under review. Once we ascertain what is actually required,  we will update this page, and the labs accordingly.

`etcd` is a distributed key-value store, kind of similar to a NoSQL database. It is the database backend chosen by the Kubernetes project for the storage of cluster state. It is a separate open source project that is not maintained by the Kubernetes developers, but no doubt they have some input on its development.

* [What is ETCDCTL_API=3 all about?](#what-is-etcdctlapi3-all-about)
* [When do I use --endpoints?](#when-do-i-use---endpoints)
* [What are all the certificates used for?](#what-are-all-the-certificates-used-for)
* [Should I stop API server and/or etcd during backup or restore?](#should-i-stop-api-server-andor-etcd-during-backup-or-restore)
* [How do I make a backup?](#how-do-i-make-a-backup)
* [How do I restore a backup?](#how-do-i-restore-a-backup)
    * [kubeadm clusters with etcd running as a pod](#kubeadm-clusters-with-etcd-running-as-a-pod)
      * [Troubleshooting after the restore](#troubleshooting-after-the-restore)
* [Clusters with external etcd](#clusters-with-external-etcd)
    * [Single etcd process](#single-etcd-process)
    * [Multiple etcd processes](#multiple-etcd-processes)
* [Advanced Usage](#advanced-usage)
    * [How do I read data directly from etcd?](#how-do-i-read-data-directly-from-etcd)
    * [How do I encrypt secrets at rest?](#how-do-i-encrypt-secrets-at-rest)

This file has been truncated. show original

And

have different answers, please assist.

rob_kodekloud · July 9, 2024, 8:28pm

The Community FAQ version was edited rather recently, so that’s the one to bet by. But there are slight variations as to what will work; try one way and see how it goes.

Alistair_KodeKloud · July 10, 2024, 5:54pm

If you’re worried about whether or not to stop control plane processes, for these clusters don’t worry about it.

When there is only one control plane and one etcd, then nothing untoward is going to happen. Recall that the only components allowed to talk to etcd are API server and other instances of etcd, therefore in a HA cluster you need to stop everything so that other etcd’s don’t modify data while you are restoring…

With a single node

Run etcdctl restore - this creates new files from the backup. The existing etcd knows nothing of these files, until…
You edit the etcd manifest to update the volume mount to the location of the restored backup.
You save the edit. etcd restarts, then API server restarts, then controller-manager sees the state of the cluster has changed, so deploys/undeploys things as needed to get to the state of the restored backup - all good.

Just do the restore lab a few times. What works there will work in the exam.

rahulshettigar13 · July 15, 2024, 5:00am

Thank you @Alistair_KodeKloud. This helped me to answer the question correctly and clear my CKA exam✌️