Multipath driver support for qlogic 23xx

[uwuertz]

We're installing a new xsan-environment (12 xraids, 2 mdc with osx, 11 clients under windows 2003/sp1 terminalserver enterprise edition) with vicom - san mirroring.

2 vicom engines are building a cluster; each cluster has two separate zones (redundant) on the fabric. Each mdc is located in both zones; working fine. If one zone fails, the mdc is using the other for meta/data and so on.

If we put the windows 2003 servers in each zone (hp blade 25, qlogic 23xx - card with 2 outputs) the snfs is running crazy. That means, we can see the volumes (cvlabel -l), we can mount it, but we can't access it (I/O - error and on explorer we don't see the volume - size).

If we put both cables from the blade in one zone (equal using zone 1 or 2) it's working. But we want to make sure, that if one zone fails, the communication is routed over the second.

We have tried a lot of qlogic - drivers; no changes. Has anybody an idea, what kind of driver (qlogic or hp) we should use for multipath -access (HBA-side) or (second best) for failover (one active/one stand-by) ???

Is it senseful, to use hp - securepath ? We never had this product before, and we don't know how to use this in a xsan - environment...

Any idea is welcome !

[uwuertz]

It seems that the problem is solved.

We used the qlogic - qlc-driver for the blades and it's working with active/stand-by. Multipath is not working; but we're still testing.

[aaron]

Sounds like a very interesting config (we have been playing with a Vmirror box, too). Can you post a diagram?
_________________
Aaron Freimark
http://www.tekserve.com/vcard/af.vcf

[uwuertz]

Yes, of course. I will draw one and post it. Our business is hosting; the system is used for 2 customers with about 360 outlets (=800 Users) in germany, austria and switzerland. The project is a high-availability san for lower prices than a hp-eva or emc.

The customer has made a test for calculating inventory (counting articles at the end of business year) on the new system (it's productiv with 6 XRaids at the moment). With the "old" emc symetrix it took 4 hours for 5 outlets to prepare and 4-5 hours more to finish. With the xsan, the customer calculates 10 outlets in less than 10 minutes (!) and finished 10 outlets in about 15 minutes(!)...

Our technical side: 4 QLogic 5200/5600, 2 MDC's, 8 HP BL 35 Blades under Windows 2003; 12 XRaids and 5 VMirror - Engines (testing at the moment); running as cluster (that means 10 engines all-in-all). We're expecting 16-18 blades until 02/2007.

Actual, we're running out of ports !

What do you think: Is it possible to stack 6 switches without buying a director (f.e. Brocade 24000) ? QLogic says, that 4 switches is the usual configuration. Is there a limit (bandwith, of course...) ?

When the installation is finished, i can give you the information about the project in an article.

[aaron]

We've had five 5200s fully interconnected, no problem. QLogic notes support for up to six switches in this press release.

You should also look at the upcoming (very soon) SANbox 9000 chassis switch.
_________________
Aaron Freimark
http://www.tekserve.com/vcard/af.vcf