Score:0

Ephemeral Kubernetes nodes

cn flag

I'm running Kubernetes on a bunch of bare metal machines, with a NetApp FAS for NFS storage. The nodes are all Ubuntu 20.04 LTS, provisioned with Ansible.

Ideally, I'd like to have a disaster-tolerant environment, meaning that as long as the NetApp-to-tape backup works I can completely recover from any kind of machine or infrastructure loss (been burned once by a datacenter fire, you remember the big one).

For the applications running inside Kubernetes and the few services that run outside of it (which is just two DNS servers and an UDP service), that works fine: mount the directories that hold the data of the services via NFS (or in the case of k8s services, use Trident-backed PVCs), run a snapshot on the NetApp every 24h, that's it. Tested, works fine.

The problem is the Kubernetes nodes themselves - I don't even know on just how many different ways that thing stores its data. There's stuff in /var/lib/kubelet/, there's config and a whole PKI in /etc/kubernetes, calico's stuff in /etc/cni/net.d/calico-kubeconfig... is there any guidance on how to make a disaster-proof k8s deployment?

mangohost

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.