XSan volumes don’t mount

[garges]

I’ve got 4 XServe RAIDs and 5 XServes connected with QLogic 5200 fiber channel switches. Set up 2 volumes with XSan. The XServes mount the volumes inconsistently and one of them won’t mount any volumes at all.

Frequently after a reboot of an XServe, the XSan volumes do not appear on the desktop. I have to either restart again, or power cycle, and hope that this time they’ll mount and they usually do.

But on one of the XServes nothing I’ve done makes the volumes mount. In XSan Admin if I click the SAN, “Setup,” and “Computers” I see the XServe computer, can authenticate to it and put in a software license number that is reported valid. If I click a volume name and “Clients”, the computer appears in the list with “No Access” and “Not Mounted.” Click on the either of the “Mount” buttons and status changes briefly to “Mounting” then reverts to “Not Mounted.”

On the system that won’t work I’ve wiped the hard drive, reinstalled Mac OSX Server 10.4 from scratch, reinstalled XSan 1.4, done all updates, and even replaced the fiber channel card and cables. Nothing helps.

On the QLogic switch, the port that connects to this system lights up with a green link light, just like the working ones. In SAN Surfer the port properties of this system are identical with other working systems. I/O Stream Guard is Enabled, Device Scan is Disabled. And I’ve tried using a different port and that system still won’t mount the volumes.

If I open the Transaction Log in XSan Admin and capture during clicking “Mount” I see only one difference. For the systems that work, the status number is “0” while the one that doesn’t work has “100007”:
status
100007
type

Any ideas on what’s wrong here?

[maxroper318]

Are you getting any errors in the system.log when the system boots or attempts to mount the volume? Can you see all of the targets in your Fibre Channel listing for Apple System Profiler? Are you using local hosts files or DNS for the systems to be able to find one another on the SAN? Are you using Xsan Admin from a MDC or from the local workstation to attempt to mount the client? Are there any differences in the /Library/FileSystems/Xsan/config directories between a working host and a non-working host? Is it running 1.4.1 or 1.4?

Hopefully with these questions you might find the cause or provide more info for us to help you find the cause.

Thanks,

Max Roper
Technology Consultant

318, Inc. | Technology :: Solutions :: Consulting
http://www.318.com • mailto:[email protected]
(310) 581-9500 voice • (310) 581-9515 fax
:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:.:
http://directory.microsoft.com/MPRD/support/AboutPartners.htm
http://train.apple.com/news/success/three18.html

[garges]

Thanks for your suggestions, all good ones.

No errors in system log.
Yes I can see all the XServe RAID targets listed as XSan volumes in Disk First Aid (even on the machines that can’t connect.)
I’ve got identical /etc/hosts files on each machine that lists them all.
I’m running the XSan Admin app from the MDC
Don’t see any differences in the XSan/config
And it’s running 1.4.1

I think it may be solved though. I was talking to QLogic support and they asked me if the fibre channel card has a heatsink. I looked and no it doesn’t. They said it must be a first generation card (it is a few years old) and there were problems with those getting too hot to function. Later cards added a heatsink.

Then they suggested a workaround, that I thought was crazy but seems to have fixed things. Plug the old first gen cards into the highest numbered slots of the FC switch. They said the higher numbered ports are electrically different and are more tolerant of the malfunctions from over heated cards.

So I did it and the SAN volume which I’d never been able to mount just mounted itself automatically. Several trial shut down reboot cycles later and it works fine.

[Author]

Posts