I'm really interested in how different people manage backups for their control plane. Do you typically opt for regular etcd snapshots, complete VM snapshots of control plane nodes, or a combination of both?
5 Answers
I use Velero for backups. It works well for saving the state of your clusters.
Backing up etcd seems a bit wild to me, honestly. I prefer relying on GitOps.
I mix etcd snapshots with ZFS for send/receive operations. Works like a charm!
I don’t actually do backups since everything I run is immutable. I keep stateful components outside of Kubernetes by using DaaS. If a serious failure happens, I’d just spin up a new cluster. It really depends on your specific use case, though.
Exactly! Infrastructure as code is a must, right?
I'm into Velero too, but I also take etcd snapshots. It’s a solid strategy!
Totally agree with you! Workloads should be backed up separately, and clusters can be treated as disposable. Using GitOps for managing deployments is key.