Solaris

Has anyone ever relocated an HBA card after already having it configured in another PCI slot? I've found that the HBA instance #'s start incrementing past the previous. It results in me having to have my instance numbers in /kernel/drv/qla2300.conf be 4,5,6,7 instead of 0,1,2,3. Cleanup with 'devfsadm -C' and a reconfigure on reboot are no help.

Thanks much,

Keith

I have never done that, but I would try "update_drv -d" on it. We use emulex hba's and they come with a lputil menu program. I might also poke around there.

Why are you moving the hba? I have heard that some pci slots may be faster than other pci slots but I don't know how to id the "fast" ones if indeed this rumor is true.

Thanks for replying. Yes, I had found that Qlogic has a similar utility and I can address it there. I'll post more details and the full solution later.

Like most rework - this was caused by someone not following instructions. It isn't that I'm seeking better performance (and I'm very doubtful of the magical slot myth). The client I'm doing work for has a pre-production lab that is supposed to be identical to production in every way. Someone built the system and seems to have randomly put cards in varying slots and PCI bus(s). The VxVM/VCS/SCSI cleanup was fun as well.

Keith

You didn't say what the server/workstation model is. If it's a Sun server like a E220R or E420R, then yes, the id's will change if you move the hba. Look for your server manual and refer to the pci slots on the motherboard. The Sun System Handbook site, http://sunsolve.sun.com/handbook_pub/Systems/ , has diagrams of the motherboards.
Some of the pci slots are 5v, 3.3v or a 3.3v/5v (meaning either voltage pci module will work).
I know that some fiber channel hba's will keep their lun id's when an extra hba card is installed, but you'll have to refer to the hba manual to be sure.

good luck.

kduffin and I were in error. antalexi may understand the truth, but if so, he didn't explain it very well.

Starting with antalexi's link, I clicked on V880 and then "Hardware Specifications" bringing me to http://sunsolve.sun.com/handbook_pub...ared/spec.html where I see:

Code:

Nine full sized, hot-swap, PCI slots compliant with PCI specification Revision 2.1:

    * Seven slots operating at 33MHz, 64 or 32 bits wide, 5 volts
    * Two slots (slots 7 and 8) operating at 33 or 66MHz, 64 or 32 bits wide, 3.3 volts

So indeed, on this server, two PCI slots are twice as fast as the rest.

Yes, you can find the 66MHz slots on sunsolve.

Also, what are you using for volume management? If using Veritas Volume Manager, there is no need to worry about the controller numbers changing.

If using Sun Disk Suite, I think you will have problems.