I'm designing a little vSphere 5 cluster for one of our remote sites. We have some IBM x3650s that have 6x 300GB 10K RPM drives in them, along with dual quad core CPUs and 24GB RAM. Because we use HP P4500 G2s at our primary site, we have licenses available for HP P4000 VSAs. I thought that this would be the perfect opportunity to use them.
Below is a basic drawing of what I want to accomplish:
I want to run a P4000 VSA on each server and run them in a Network RAID-10 (Lefthand speak for network mirroring, think of it as RAID 1 across nodes or as an active/active storage cluster). I will then present this storage to guests that will run on this mini-cluster. It will be managed by a vCenter Server on our main site.
All connections will be GbE with two dedicated to storage. Management and Data will share a pair of connections, since I don't expect there to be high load. These servers are just there to provide directory services, dhcp, printing, etc.
Does anyone see anything potentially wrong with this approach? Is this the best way to do this without adding additional dedicated storage heads? Are there any pitfalls in this design, besides the lack of dedicated Data/Mgmt interfaces?
This looks like perfectly reasonable design to me. I assume you want to place your virtual machines on data stores on VSA-based HP LeftHand Storage. You have the hardware and networking part covered: using two NICs on vSwitches for your HP LeftHand VSA, the ESXi's iSCSI VMkernels and possibly guests which access the HP LeftHand cluster is best practice; I suggest you create resource group reservations for the VSAs according to the user manual.
[Edit]Because just saw this in the comments to the original question: It is best practice to presented RAID-ed storage to the VSA which is made then virtualizied and made available in the HP LeftHand Cluster. The RAID level is depended on your requirements for capacity, performance and protection. RAID 10 is the way to go for you then.[/Edit]
One thing you need to be careful about is where you place your managers in this HP LeftHand setup. This is very important in a configuration with only two storage nodes! Right now I do not see a HP LeftHand Failover Manager (definitely the preferred option) in your design to maintain quorum in the storage cluster; or are you planning use the Virtual Manager? Depending on the uplink to your main data center (latency <= 20ms, bandwidth ~100Mbit/s), you might be able to place it in there.
PS: You might also want to make use of Remote Copy to replicate your VMs and data into your main data center; you already have everything which is needed in place to make it work.