Take the 2-minute tour ×
Server Fault is a question and answer site for professional system and network administrators. It's 100% free, no registration required.

Gents, what would you recommend for production use in a FC environment, where live migration is a must?

Which clustered file system should we use? OCFS2?

Would you still need drbd?

Thanks!

share|improve this question

3 Answers 3

up vote 1 down vote accepted

We've used OCFS2 very successfully for a while now, I know there may be bigger/better filesystems out there but I trust OCFS2.

share|improve this answer
    
thanks for your help! How does the setup look? Fibre channel storage attached to your dom0s, then OCFS2 file system? –  CMag Jun 20 '11 at 19:35
    
I'd be happy with that kind of setup yes. –  Chopper3 Jun 20 '11 at 19:50

Remember that you don't need a cluster filesystem if you don't use imagefiles. In most cases, it's far faster (and lighter on the host/Dom0 cache) to use block devices (LVs, LUNs). If you go that route, you only need cLVM or plain LUN management.

share|improve this answer
    
high five, for sure ! –  CMag Aug 15 '11 at 16:01
    
With SAN I see no need for DRBD or a clustered filesystem in that use-case, too. –  Nils Aug 30 '11 at 20:15
    
@Nils: that's what i meant by LUN management (which is a feature of the storage target, not the SAN). As for DRBD: you're right, if your array controller has enough redundancy. –  Javier Aug 31 '11 at 1:59

Whether or not you need drbd really depends more on your requirements for data availability than on the technical details of your storage connectivity.

Arguably in most environments you don't "need" drbd; you use it to provide some level of data availability that would be difficult to achieve without it. The one exception is if you're using drbd's dual-master mode to emulate shared storage...but you say you are in an "FC environment", which I'm going to assume means your virtualization hosts have access to shared storage. Even in a SAN environment, you may end up using drbd as a remote replication solution.

You haven't provided a lot of details about your current environment, so it might be difficult to provide good recommendations. For smaller clusters, RedHat's GFS might be a good option -- if you're running on RedHat/CentOS/Fedora hosts, the packages are already available as part of the stock distribution. People also seem to be having luck with OCFS2, and the GlusterFS folks also claim their product works well for virtualization storage.

Depending on your environment, simply setting up a NFS server and accessing storage via NFS might provide acceptable performance -- and would be substantially simpler to set up than any of the clustered filesystem options.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.