This may be more of a physiological question but I am curious what other storage administrators think, and what they are doing at their companies. I am in a new environment where NFS is the standard storage protocol for VMware. Nothing wrong with this IMO, as this is a VMware endorsed way of presenting storage to ESX hosts. My question is about further presenting storage via NFS/SMB mounting in the guest. It seems to be a huge thing in the environment I am in, as well as in-guest iSCSI. I have always presented storage to ESX, either through FC, iSCSI, or NFS, and let it dole out the storage through datastores. This feels more clean and less complex to me as you aren't mixing and matching storage access. Most VM storage is presented as datastores, but the larger systems (>2TB) have large NFS/SMB mounts to other NAS storage outside of VMware datastores here, which are serving up data to the app the VM is running. Does anyone here present storage this way? I just want to see if anyone has an opinion on this based on their past or present experiences.
In terms of manageability, it's obviously better to consolidate all external storage devices under single (or multiple if we are talking HA) "head" server.
Direct "NAS to Client" storage provisioning can be considered for small-sized deployments. For example, you can provision SMB shares from your local QNAP box to your office workstations and it's totally fine.
Provisioning NFS shares directly to the virtual machine may not be the best approach in your case. Instead of having one "entry point" which is your ESXi host or HA Cluster, you will get two.
I would recommend you presenting the external NAS to your ESXi as NFS datastore and provisioning storage for your VMs as .vmdk drives instead of creating shares.