This is a complex question, but lets say you want redundent pathing, so 2 controllers, 2 switches, each ESX host has two HBA ports, this is a common model. The dual-controllers gives you the cross-over effect you want to have redundent pathing to all LUNs as defined on the ESX host. HP MSA 1000, 1500, etc. can do active/active, but VMware does not support this consistently, or at least when we setup our default configuration initially. So 4.48 is preferred by us, stable and known animal, have tested 5.20 but found some quirks.
Zoning, must setup zoning per switch, this is the switch per MSA controller, think of them as a matched set, the Controller interfaces to the Switch, but they are separate devices. Must set the profile of the interface on each switch per HP guidelines. Must setup SSP in ACU per HP guidelines, per HBA port. HP has a rather good PDF on MSA1000, 1500, etc. setup for VMware. I can dig it up, but a search on the HP website should find it without issue
You really want to watch the total Disk IO loading on MSA, especially the 1000. We have flooded MSA1000s with only 4 ESX hosts, and you will see some signfiicant differences between RAID 1 versus RAID 5 sets.
MSA1000 has 4 key areas that must be configured, ACU GUI (DASD setup and SSP), brocade switches via CLI and via GUI for Zoning (make sure to enable the switch configuration once created), and MSA Controllers via CLI to setup IP (then use GUI via web/http). The HP instructions are a bit confusing between MSA Controller CLI and Switch CLI, and you will need serial connectivity to setup the switch and controller IP addresses, but once you have IP addresses assigned to the MSA controllers and switches about 90% is web or web/java GUI based.
One last point, that VMware and HP are not clear on, if you plan to use VCB you MUST setup explicit zoning on the switches. If you only setup SSP, and use the default zoning VCB will NOT work right, spent 6 hours on the phone with VMware only to be told the default zoning is not handled by VMware VCB code right. This has not, per my initial review, been fixed with VCB 1.1, but sense we do not use VCB any more, I have not revisited this issue recently.
Now, for the ugly issue... You must have at least 1 Windows based HP server connected to the MSA at all times to monitor the MSA correctly. We actually use a Windows based server for each 2 or 4 MSA1000/1500s (connected via fiber HBA port, we use older DL380s with HBAs), if you do NOT do this, you will miss critical hardware alerts. HP does not explain this well. Even if you have HP servers runing ESX OS, you STILL do not get complete fabric monitoring or alerts without the Windows server connected via fiber to MSA. Never mind the fact, that that the GUI based ACU and ADU tools are ONLY Windows based and only communicate via fiber link to MSA. You don't want to hear the ugly things I said to VMware and HP but this issue not being explained effectively on the respective HCLs... I said things in German, not English... I only do that when I am really really not happy with vendors. LOL
