MSA1500 - Controller failover causing issues with MSCS

[UTryIt2]

Greetings,

Recently our company has started having problems with our MSA1500 and the 2 / 2 Node Windows 2003 R2 Clusters.
Also attached to the MSA are 1 VMware ESX Server and 1 MS File server. For some reason the MSA fails from the active controller to the passive one and causes our MS systems to loose their storage thus causing our clusters to start thrashing and never recover.
I also noticed with the MSA in where it does not provide a true redundant architecture. I'll try and draw what we have below:

|--A--| |--B--| One Cluster Group
|\ /|
| \ / |
| \ / |
| /\ |
| / \ |
|/ \|
(--1--) (--2--) Fibre Switches
\ /
| |
| |
a| |b
[--MSA--] HP MSA1500


First on the redundancy, if controller (a) is the active controller on the MSA, and I loose switch 1, I loose all storage on my systems.
The MSA controllers don't support dual fiber cards, which is concerning.

In a test of the MSA pulling the cable from node A to switch 1, causes the MSA to fail over after some time, but in that delay, all systems attached to the MSA have lost their storage.
The only system that does not fail is the ESX server.

Additional Information:
The MSA is at the most current firmware.
The QLogic HBA's are at the most current firmware
The Servers (HP Proliant DL380 G4) are at the most current firmware and drivers
The MCDATA fibre switches are at the most current firmware
The ACU software is at the most current version.
We are using StorPort Drivers and MS MPIO Drivers.

Will placing the MSA in an Active / Active configuration help resolve my redundancy issues?
Has anyone else experienced similar issues?

Any advice would greatly be appreciated.

TIA,

Gus

[Uwe Zessin]

That looks like a configuration problem to me.

Have you made sure that each server has a connection-entry on each controller and that its operating system profile is set correctly

Is SSP set up properly? Did you verify in the Qlogic adapters if they can 'see' the MSA disks properly?

[John Kufrovich]

Gus,
UWE could be correct.

Based off your diagram, you have MSA C1 as the left side. Actually MSA C1 is the right side.

With VMware, be aware of your SAN configuration. Steps to check.

1)Check your PCI slot assignment. Some slots share a pci bus. With some OS, slots have precedence.

2). If the HBA is dual ported, Port 0 has precedence over Port 1.

3). "Square Up" not a technical term. Make sure you have either the first slot or p0 going to switch 1 -> MSA (right) Controller 1.


I looked up your servers slots(Dl380G4), here is the order. Looking from the back.
3
2
1

Slots 2/3 are shared, so that bus will only run as fast as the slowest PCI card.