conflicting addresses on fc loop

[vladimir_1]

N4000 server has two A5158A adapters and works via FC hubs with Symmetrix. After adding third such adapter connected to Ultrium library via a 2400 switch and reboot the server complains: "It appears that there are more than one device with the same address (nportID) setting of 0x0. One of them will be ignored". Where are conflicting devices and how to overcome this problem?

Thanks.

[David Meidam]

How many disk arrays are you connected to?
What type of switches/hubs are you using?
What are the results from:
fcmsutil /dev/tdX devstat all | grep Nport

Does resetting the card produce anything different?
fcmsutil /dev/tdX reset

Try disabling the card, reset the hub, re-enable the card. Does this assign a different N-port ID?

[Michael Tully]

Hi,

Make sure that you have patch PHKL_21834 and
the latest driver for A5158A (B.11.00.06) loaded.
If you have a Tape library unit located on the
same switch you will also need:
PHKL_23790
PHKL_22759
PHSS_24044
and the kernel parameter 'st_ats_enabled' value
to be "1".

Are using the switch port as fabric or arbitrated loop?

HTH
-Michael

[vladimir_1]

Sorry for I gave to little info. First two adapters connected via A3724A hubs form an arbitrated loops. Third adapter connected via brocade switch to Ultrium library and I suppose should form a point to piont fabric connection. Do I have to configure the switch for such mode or it does this automatically.
At this time tdutil for first two adapters shows NportID 0x00001 and 0x00002. And for the third one 0x00004. But the hubs and the switch are not connected with each other so the ultrium port is the only port (besides HBA's port) connected to the switch. Hwo can it conflict with? Patch PHKL_21834 and latest drivers are installed.

[Oliver Haensel]

Hello Vladimir,

the first 2 HBAs are in a seperated loop from the 3rd one. That's why they don't affect the new HBA. So focus on troubleshooting the new HBA.
tdutil tells you NPort ID 0x4 -> This means you're running in private loop mode. Probably you've set the switch port to the server into quickloop mode. I would recommend to disable quickloop (qlportdisable command) on both the host and library port! The only exception should be if an old Tachyon HBA is used which is not yet fabric capable.

-Oliver-

[Oliver Haensel]

one addition: the error message you get is normal when you are connecting 2 switch ports to the same hub (or 2 connected hubs) as the switch port always gets ALPA 0x0 in the local loop which has the highest priority. If you connect more then one switch port to a loop all switch ports but one will be deactivated.
(Quickloop is a feature to form a single loop consisting of all local loop(let)s)

[vladimir_1]

Hello Oliver,

The switch is in default state but I don't know how to see which mode its ports are in and can't find qlportdisable command in the list of switch OS commands. Is it for this type of switch?

Vlad

[Oliver Haensel]

Hi Vladimir,

you've said it's a Brocade Silkworm 2400 - 8 fc ports and no front panel.
You must telnet to the switch or establish a serial connection to it (Default: username: admin, password: password)
qlshow tells whether or not the ports are in quickloop mode (if you have the quickloop license installed, you can check it with the licenseshow command)
switchshow tells you where the ports are connected - and in what topology

[Steve Bonds]

A very useful QuickLoop command is:

qlhelp

which will let you know all the other Quickloop commands.

Before changing your QuickLoop configuration you should read the Brocade QuickLoop manual and double check that your HBA is not hardcoded to fabric mode. (If the HBA is fabric-only, enabling QuickLoop will effectively bring the port down.)

The HP/Brocade QuickLoop manual is at:

http://www.hp.com/products1/storage/san/fibreswitches/silkworm2800/infolibrary/index.html

Are you having connectivity problems, or just concerned about the message?

[vladimir_1]

And a final question. How could I change a al_pa of a HBA if needed?

Ths.

[Vincent Fleming]

The AL_PA is not set on the HBA - it's set on the storage devices (ie: targets). I think you need to set the AL_PA on the tape unit; but I can't tell you how (I know how to do it on an XP... but that won't help you.)

Good luck

[Steve Bonds]

You should avoid hardcoding an AL_PA anywhere if at all possible. The fibre channel protocol has a rather thorough method of self-assigning AL_PA addresses. Locking a device to an AL_PA creates the potential for duplicates (if two devices get hardcoded to the same AL_PA) during the next loop initialization. Even if there are no duplicates, it makes the loop initialization more complex than it needs to be.

[Doug Rauenzahn]

You should definitely be using fabric (n-port) and not Quickloop or Public loop (NL- ports) for the Ultrium library. I'm assuming you have FC-SCSi bridges and an HP Ultrium library, telnet into those bridges and set them to soft-addressing mode. If your bridge has a web browser interface, set the bridge to n-port mode.

Also, if you plan on using multiple hosts on the SAN, definitely set st_ats_enabled to 0. That kernel parameter should only be set to 1 in a shared environment if you are using ATS services packaged with MC/Serviceguard which doesn't support switches.