Well, I've been unable to collect good answers on this by using Google, so I opened a case with VMWare, but would also like some community opinions. We're looking at using dedicated Linux boxes as NFS servers, not anything like a NetApp due to cost constraints. We're on ESXi 4, by the way.
Here's my question list. Please, if I've overlooked anything don't hesitate to point it out to me:
What mount options are used internally by the hypervisor when mounting via NFS?
What NFS configuration options are recommended for best performance & data integrity? Is async safe provided our NFS servers have adequate power protection? Any thoughts on wdelay vs. no_wdelay?
For mounting the underlying filesystem, any suggestions on FS type, block sizing or other tuning that is known to work well for large (>1TB) datastores on NFS under ESX?
Is it safe to mount with noatime? relatime? (aka: Does ESX rely on the "access time" attribute of files for anything, or just mtime & locking?)
