HP Lefthand P4000 - What challenges did you face?

[hutchingsp]

OK so our new SAN search is nearing an end, and as of right now the HP P4000 is looking like the most likely solution as it offers the capacity and performance we want as well as seeming to offer some very nice HA features without breaking the bank.

Those of you who have one, particularly in a multi-site configuration (or a single P4000 site split across your physical sites), what problems did you encounter please?

The (rough) plan would be 2 locations, switches in each linked by direct fibre so very low latency, nodes in both sites, vSphere host(s) in each site, HA certainly, FT on key VM's ideally (maybe not from day 1) and the theory is lose one site and the other takes over the VM's with minimal interruption as the storage should be available.

Thanks in advance.

[dcoz]

There are a number of things to be aware of.

The first and pretty important points is the locatin of a third storage node or FOM. Ideally you want to have a physical third location so if one storage node goes you can still maintain quorum.

Second is you need to have adaquate bandwidth. It is basically 1 Gbps of bandwidth for two nodes, and 2 Gbps for 4 nodes. To help and keep storage traffic to a particular site have a look at this PDF http://h20000.www2.hp.com/bizsupport/TechSupport/CoreRedirect.jsp?redirectReason=DocIndexPDF&prodSer...

Be aware the PSP of round robin isnt supported in this scenario.

I would also have a look at this blog post http://thehyperadvisor.com/?p=656

Third is really understand HA if you are going to split your ESX HA cluster. I have seen an issue where there was HA split brain when the link between the two sites was lost. The next point related to this is think about where your gateway is and if you lose a site how you will deal with this.

There are a number of things to think about but hopefully this will get you started.

Dougie

[JDLangdon]

I just setup some HP Lefthand P4000 virtual appliances in our lab and honestly, the hardest chllenge I faced was trying to find the correct version of the CMC to manager them with.

________________________________

Jason D. Langdon

[hutchingsp]

Thanks both.

The (initial) plan would be two sites, not sure of node numbers yet. From digging around in a twin site config the FOM would go in the site we consider our primary, which fits in with how we view both locations.

Bandwidth, we have a 12 or 24 (I forget) fibre with most of the cores spare so we can drive it at 1gbps or 10gbps.

Tbh the P4000 "split brain" stuff seems simple as I've tested it with the kit we have on loan, it's how vSphere will cope if the link goes down and the FOM kicks in to take the storage offline at the "B" site - HP's PDF's seem to concentrate on FT but don't say much about HA.

Appreciate any light you can shed.

[hutchingsp]

Hmm ok so it seems that the P4000 multi-site configuration is to have separate subnets for each site.

It also seems that the vSphere iSCSI initiator doesn't support destination that aren't on the same subnet.

So those of you who have multi-site setups with vSphere how do you handle this please?

The best solution so far seems to be to have one VLAN per subnet, have the VLAN span the switches (fast link) and have multiple vmkernels per host with vmkernels across both subnets/vlan's on each host, but apparently even this can mean some manual intervention if storage goes down in one site and a host has LUNs bound to a VIP that isn't responding (not a showstopper just trying to understand it).

This side of things doesn't seem all that well documented by HP.