41. ● Test that everything can be deployed on new version
Test deployment
42. ✓ All basic Kubernetes resources must survive upgrades
✓ CRDs that you use must survive as well
✓ Write sample app to measure downtime
✓ Check that new applications can be deployed on the cluster
✓ Check that settings haven’t changed
Test upgrade
43. Recap
How to upgrade
1. Check the cluster health
2. Backup
3. Upgrade etcd
4. Upgrade master
5. Upgrade worker
6. Check the cluster health
Keep the applications
running
● Make applications HA
● Cordon and drain nodes
● It takes time
Prepare to the upgrade
● Test deployability of a new
cluster
● Test upgrades
44. ● Automate upgrades
● Do maintenance during working hours
● Don’t write your own installer
Bonus