Are people leaning more toward RDM or VMFS these days? What are YOU running and why?

[LiamCurtis]

Hi Everyone,

I know there is a wealth of info on the topic and the standard answer is to use both...whether you want array or vm level snapshots, etc.

However, my environment is staged, I've done some torture tests, backup and restore tests. I'm almost ready to go p2v wild, but I'm still on the fence about where to use RDM vs VMFS. I am running a cx500 and I was bitten by a flare code bug a while ago that caused unnoticed corruption in my BCV clones that we were using for backups--this was happening FOR MONTHS until we discovered it after we ran into weird anomolies with a cloned Oracle instance. From that point on I never trusted snaps or clones on the array side. On the other hand, I do like the quiescing capabilities of a vmdk.

Seems like there are some great advantages to going VMFS especially with VDR and similar tools sprouting up. BUT..there is still that weird feeling about relying on those vmdk files. Just curious what percentage of VMFS vs RDM people are running and why you personally and absolutely chose RDM vs vmfs.

Thanks in advance.

[RParker]

RDM was good in the past, for certain products that required RDM access, like SAN tools, exchange and clustering.

But those things are becoming more mainstream on VM, and VM Ware has made VMFS for friendly. So RDM need is waning.

You can cluster without RDM now, besides which VM Ware has their own fault tolerance, but we run 100% VMFS, don't have a single RDM now. RDM is a hassle and VMFS / vmdk is much easier to manage.

[LiamCurtis]

Thanks RParker. I do agree...seems harder to keep track of where all the RDM's live from within VC...

+1 for VMFS

[DSeaman]

Doesn't Windows server 200x need a RDM for the quorum disk?

[Josh26]

As soon as you implement VMWare, someone will decide "it should be trivial to move storage to other datastores, etc". I got sick of answering this, and will never look back at RDM. Note I'm not commenting on what might be technically better.

[jayctd · Accepted solution]

As we are ISCSI in the 3.5 world we direct attached with MSISCSI tot he virtuals, this was for performance reasons as 3.5 did not have MPIO

In the 4.0 world we are leaning towards VMFS the feature set and behavior in a SAN interruption are advantageous to us and now with MPIO we are seeing great performance out of VMFS's

Jered Rassier

*EqualLogic Technical certified professional

*Dell Enterprise Foundations v.2 Certified professional

##If you have found my post has answered your question or helpful please mark it as such##

[LiamCurtis]

Thanks for the replies, all!

Yeah...I think I'll stick with VMFS and VCB. I think advantages on the virtual side outweigh using RDM.