High Availability Hyper-V Clusters

[Mr.Ben]

Good afternoon all,

 

As part of the brief for my new job I'm looking at Hyper-V Clustering to reduce the amount of Servers that the school currently has (there are 28, 20 of which are possible targets for Virtualising).

 

As I'm researching this I'm looking for a school with a Hyper-V Cluster to speak to to find out what they feel the drawbacks/ advantages to such a system are, as well as how they went about it and if there is anything that I should be watching out for.

 

Thanks in Advance, Ben

Edited May 18, 2012 by Mr.Ben
Awful Splelgin :-)

[strawberry]

Hyper V clusters are great, you can migrate servers in real time from one host to another. It makes excellent use of Ram and CPU speeds as you can buy servers with very little in the way of disks. It does however mean that the SAN you use is the central point of failure. The key to it is to duplicate servers, use DFS, clustering and multiple servers spread accross different hosts. Some using clustering ,some standalone.

[sonofsanta]

I'm just setting ours up here - seriously, watching a guest live migrate from one set of hardware to another without dropping a ping is magical - and it seems pretty stable and do-able to me. It's difficult to make full use of all your resources with it, but you can get around it with a bit of cleverness - for RAM, you can set a minimum amount on each guest at a level where all guests could co-exist on a single server, but you can set a dynamic amount that allows it to expand when hardware allows (i.e. both hosts are up). You can't, unfortunately, do the same with CPU cores, but my approach to that has been to treat the Xeons as double what they are, to take into account the hyper-threading. That way, every physical core can be used when everything is fine, but when one host is down, there's still sufficient logical cores to handle everything.

 

It takes a bit of work to wrap your head around it all early on - documentation online can be, er, fragmented - but once you dive in it's actually pretty straightforward. I was keeping track of it all on my blog here but then got busy actually doing it instead of just writing about it.

 

As @strawberry says, your SAN becomes the heart of everything, but most SANs should have built in redundancy anyway - dual controllers, dual PSU, redundant cabling etc.