Storage servers should get more block devices than test runners #28
Description
Some cloud providers limit the amount of block storage that can be attached to a set of instances in one region. For example Google limits the amount of block storage to 500GB for anything faster than the bottom end block device.
This is fine for workflows with a single instance. For NFS, however, each instance gets the same number of block devices, and often we need several instances (an NFS server and one NFS client for each NFS version, say).
For the NFS server to have enough block storage to handle all the provisioned NFS clients, each /client/ also gets the same amount of block storage, but leaves it unused.
Unfortunately that means for the fstests workflow, where each NFS client needs 50GB of storage on the NFS server (for the test and sparse partitions), there isn't enough block storage in the 500GB limit to accommodate more than a single NFS client.
One way to address that is to provision 32GB (or less) for each NFS client, but then enable the NFS server to get more block storage than the test runners.