Backup strategy
- Docker & Kubernetes course
- Expert
- Lesson 37/47
- 4 min
Learn cluster backup strategies.
Backup strategy
To recover a cluster from disaster you need to know what is backed up. There are three layers:
- Cluster state: all Kubernetes objects; it lives in etcd.
- Persistent data: the database and files on PersistentVolumes; it is not in etcd.
- Definitions: the manifests, charts and values files in Git (already versioned with GitOps).
The 3-2-1 rule: 3 copies of the data, on 2 different media, 1 of them elsewhere (another region or account).
etcd snapshot -> s3://yedek-eu/etcd/ (every hour)
Velero backup -> s3://yedek-tr/velero/ (every night, keep 30 days)
Git repository -> GitHub + mirror (every commit)Testing the backup
A backup that has never been restored doesn't count as a backup. A restore drill to an empty cluster is done regularly and the duration is measured. Encryption and retention of backups are also tied to policy:
# every month: drill
1. set up the staging-dr cluster
2. restore the latest Velero backup
3. run the application tests
4. record the duration (target: under 1 hour)Practice List the backups
Write three commands: list the files under s3://yedek-eu/etcd/ (aws s3 ls); list the PersistentVolumeClaims in the cluster across all namespaces (-A); list the Velero backups (velero backup get).
Starter code
# buraya yazSample solution
aws s3 ls s3://yedek-eu/etcd/
kubectl get pvc -A
velero backup getExpected output
2026-11-03 09:00:12 48234112 etcd-2026-11-03-0900.db 2026-11-03 10:00:09 48301660 etcd-2026-11-03-1000.db NAMESPACE NAME STATUS VOLUME CAPACITY ACCESS MODES magaza veri-db-0 Bound pvc-7a1c 20Gi RWO NAME STATUS CREATED EXPIRES gece-2026-11 Completed 2026-11-03 02:00:00 +0300 +03 29d
Questions and answers
Review the lesson questions together with their answers. To test yourself, solve the lesson interactively.
1 In a cluster where only etcd is backed up, what is lost?
- Deployment definitions
- ConfigMaps
- Secrets
- The database data on PersistentVolumes Correct answer
Explanation: etcd holds the objects; data on disks must be backed up separately.
2 The number of copies kept elsewhere in the 3-2-1 rule.
3 copies, 2 media, ____ of them in another region3 copies, 2 media, 1 of them in another regionExplanation: One copy must be where the disaster can't reach.
3 Match.
| Term | Meaning |
|---|---|
etcd |
Cluster objects |
PersistentVolume |
Application data |
Git |
Definitions |
tatbikat |
Restore trial |