SAN boot with QLOGIC QMH2462 HBA adapters - Question

[Rob_69_1]

Hi all, my system is composed by:
- #1 C3000 Enclosure
- #2 BL460C Blades
with Qlogic QMH2462 FC HBA onboard
- #2 Brocade 4/12 SAN Switches
- #1 DOUBLE FC CONTROLLER 2012FC disk array

The HBA's are both on Mezzanine 2, so they map a port on each of the FC switches.

The switch zoning and cabling should allow each HBA to see the proper "boot" LUN for each blade.

In a matter of fact, when I try to configure the "selectable boot" option on the HBA's, I see 2 different WWNs pointing to a "MSA2000" device (I believe they are the 2 controllers?).

BUT JUST SELECTING THE FIRST WWN PRESENTED IS ALLOWED.

If I select (as secondary boot option) the second WWN, a "selected device is not disk device" error message is thrown.

What am I doing wrong? I'd like to be able to boot from the 2nd switch/controller if the first path wouldn't be available for any reasons!

Any help/idea would be greatly appreciated...

Thank you

[Rob_69_1]

Sorry, I realized I was a bit tired when I wrote down the original post (I see it's very confusing). I try to be more clear.

I have a 2012FC double controller Disk Array.

I set up the 2 controllers FC topology as "point to point" (as ports are connected to FC switches).

The 4 Ports on the 2 Controllers are NOT interconnected (same reason as above for this configuration).

I've properly mapped/zoned a host with QLOGIC QMH2462 HBA onboard, that has to boot from SAN (and IT ALREADY DOES BOOT FROM SAN CORRECTLY).

The problem is: when I configure the QLOGIC HBA, I'd like to "tell" him to eventually boot from Controller B, if Controller A is not available.

So, I go into the "selectable boot settings" on the QMH2462 Setup Utility (CTRL+Q at boot time).

There, I SEE THE WWNs OF THE 2 CONTROLLERS on the 2012FC (more precisely, I see the WWN of the Port were the FC cable is physically connected). But if I select to boot from a Port BELONGING TO CONTROLLER B, a "selected device is not disk device" error message is thrown.

LUNs are properly mapped, and I repeat, if in the "selectable boot settings" I put in a WWN belonging to Controller A, it DOES boot correctly.

What I want to do is to put in a "second choice boot path", in case Controller A wouldn't be available.

Am I missing some basic concepts and therefore doing a mistake somewhere? Is it a known behaviour for the 2012FC to act like that?

I thank you in anticipation for any clue/hints, that would be appreciated...

PS: I already tried ALL the combinations with Loop Topology too. Same behaviour...

[Uwe Zessin · Accepted solution]

The MSA2000 is a crazy box, ahem.
In the Generation 1 (=MSA2012), one controller manages a virtual disk and the controller presents all its volumes. The other controller DOES NOT present any of these volumes - not even inactive SCSI LUNs !!

If this controller fails, then the other controller takes ownership of the vdisk and the volumes.

Here is a customer advisory that gives some background information:
http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?lang=en&cc=us&objectID=c01693126

As usual, the formatting is awful, but maybe you can decipher it. (I've reformatted it for my own use, but don't know if I can redistribute without negative impact, sorry)

[Rob_69_1]

Thank you so much. BTW I don't know wether I should laugh or cry...

OK, "second choice boot" should be allright, as the doc you pointed me to says that in the event of a controller failover, the other controller gets the "right" WWN as well (so it' always the same WWN+LUN I have to point to, to have the system booting).

But now, a couple of questions, if you'd be so kind to help again:

1) What's the recommended config for the Controllers/Ports, then? I THOUGHT it should be OK with "point to point" for all the ports, with NO interconnection. But what about Zoning the switches, if a failover occurs and the WWNs of the 2 Controllers are magically flipping in the air and getting swapped????! Maybe FC_AL would be better? Very, very confusing!

2) How reliable can be a Multipath configuration, in case of a Controller failure? There will be a "dead zone" during the Controller Failover. What about, for instance, a Database running and "loosing connection" for a second or two?? Scary.

3) Ok, a switch failure, at least, should not cause big issues as long as Multipath has been correctly set up.

Can I have your valuable opinion regarding points 1) and 2) please?

Thank you again, and regards...

[Uwe Zessin]

1: loop (=FC_AL) without interconnect enabled. Yes, I know that point-to-point sounds more logical in a switch environment...

2: Yes, very scary. As I like to keep my forum account with write access I refrain from writing my thoughts here and give you another customer advisory - you be the judge:
http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?lang=en&cc=us&objectID=c01709613

3: correct, because both host ports from each controller are supposed to be connected to different switches.

[Rob_69_1]

Uwe, just thank you.