This content has been marked as final. Show 8 replies
budachst wrote:ASM is not supported on any of the VM servers. If you only have a single SAN, this will always be a single point of failure, regardless of what you do at the Oracle VM Server level. If you have multiple SANs, you could create repositories on each SAN, and then present multiple virtual disks to each VM and use software RAID or ASM inside the VM to provide fault tolerance across the two disks. Though, you will still have a single point-of-failure for the pool filesystem LUN itself, as that is not mirrored.
I am running an OVM cluster consisting of 6 VM servers, that currently access a FC-attached LUN. This setup does not offer any fault tolerance on the storage level.
Is ASM supported on VM servers at all - or what other options do I have for adding fault tolerance to my underlying storage?
budachst wrote:iSCSI is fully supported by Oracle VM Manager, but I'm surprised that your SAN has no built-in redundancy, i.e. some form of RAID.
so the only possibility is to have clustered storage in the background right? Maybe some ASM cluster that exports iSCSI targets instead of using FC LUNs on the VM servers - I suppose iSCSI targets are supported on VM servers?
budachst wrote:So assuming you only have a single SAN, how would ASM even help to provide redundancy in this space? I think I'm missing something here.
It does - but I do have only one RAID enclosure, where the RAID set consists of a RAID-10 set, but that wouldn't provide redundancy for, e.g. a controller failure. Although it provides good redundancy for drives failures.
He he - you aren't, of course. I am starting to think of the best way to update my storage to provide fault tolerance, since I do only have one RAID box today.
There're a couple of options and I though about ASM as the underpinnings then having each VM server access those ASM disks - probably via FC, but since ASM is not supported on the VM servers, I guess I will build some ASM cluster on OEL boxes that access my storage via FC and that share out my repository(s) via iSCSI.
Combined with LACP/port-channels on my Ciscos that should provide me a much more fault tolerant system. Not as fast as it is today, but hopefully good enough.
You could get creating and build a block-fault-tolerante storage setup..
I've used OpenFiler (HA setup, google it) that essentially creates two systems that have your data in sync on both, and a highly-available NFS mount in between the two. While Openfiler isn't something Oracle will support, NFS is.
Thanks for the heads-up. Openfiler has some interesting features, especially the ERL2 support, although I don't know if the iSCSI initiator shipped with OVM 3.0.3 is capable of using that. I will surely take a look at that solution as well. The only downside to me is, that it's yet another product in the mix and that I'd like to keep the setup as streamlined as possible in that regard.
Just did a quick search and confirmed that the iSCSI inititator available for OVM Server is not capable of ERL2, what a pity. To benefit from ERL2 one would have to compile the Core-iSCSI initiator, but that of course is not supported on OVM, even if one would manage to do it somehow…
Edited by: budachst on Apr 21, 2012 8:40 AM