Identify ideal storage solution for single-master cluster solution

[billimek]

The way I see it, there are several different options:

1. external nfs server (on proxmox) that is exposed to each node of the k8s cluster.
   • issue is that when proxmox node is down (but k8s nodes are migrated to proxmox-b, storage backend is dead
2. nfs server hosted on k8s host
   • both k8s and other nodes reference this single VM for nfs storage access
   • simple solution but requires host-level setup and k8s nodes are just clients
3. nfs server hosted on k8s host but coordinated through k8s itself
   • doesn't require host-level setup but instead all handled in k8s directly (for the most part)
4. glusterfs storage cluster residing on k8s, k8s-b, and a witness/arbiter on one of the pis
   • introduced complexity of glusterfs solution
   • need to worry about quorum unlike single-node nfs solution

[billimek]

Leaning towards #3. Not sure what the actual benefit is for #4 and why take on the additional complexity

[billimek]

#3 was too difficult to setup, so just went with #2.

on k8s:

/etc/fstab:

/dev/sdb        /srv      ext4 defaults 0 0

(where `/dev/sdb is a 64GB drive)

/etc/exports:

/srv    10.0.7.17(rw,sync,no_subtree_check,no_root_squash) 10.0.7.18(rw,sync,no_subtree_check,no_root_squash)

These are used in the nfs-client setup.