Generalized strategy for file server virtualization in Xenserver
- by Jamie
I'm not shopping as much as I'm looking for some guidance on good idea / bad idea strategies. I'm sure I'm not in the "best practices" budget range.
Currently, I have 3 dell poweredges running xenserver in a pool.
Each node has a ubuntu file server, serving about 6TB. One is the primary, the other two are rsync targets for backup. The 6TB is stored on their respective local storage disks as an LVM of 3x2tb virtual disks. The fileserver VM disks are also stored on the node local disks.
Each node also runs a smattering of light-weight VMs for web, development, windows VMs, and stuff like that. Several of those VM's disks reside on a QNAP NAS to play with live migration. These VM's are often clients of the primary file server (like all the mail, web content, user files are stored on the file server, not on the mail, web, and samba VMs).
This all works fine, and is a major step up for us. The downside is that the QNAP is a single point of failure. And the only thing the QNAP is doing is serving migratable VM images, not client data. Someday the poweredge local arrays will be full, and we will have to reinvent ourselves again.
Is it wise to have heavywieght vms (like the fileserver, with its 6+ TB disks) on a SAN or NAS?
Would it be better to keep the VMs lightweight, have the VM images on a SAN or NAS, and use 2 or more NAS act as NFS-serving file appliances? A hybrid SAN/NAS that can serve iscsi for images and NFS for the client vms?
It seems like live-magration would be a misnomer if you have to migrate a fileserver with its entire 6+ TB disk.
I recognize there are plenty of ways to skin the cat. We've already skinned it a few ways. What makes sense?