AIX

Ya no worries, I really appreciate you digging into thsi and one of the errors in the subsequent error logs is in fact FCS_ERR6 so please feel free suggest other possibilities along these lines. I googled quite extensively for all of them actually and tried quite a few things but started to ran out so decided to reach on unix.com for further suggestions. I'll try that page but I'll also take out SCST and try another targeting system this weekend to hopefully isolate the issue further. If I still get the same problem, then it would point me to hardware and optionally I will remove the Emulex cards and order some QLogic cards instead. I just would like to get more information before taking out my wallet.

Cheers,
DH

---------- Post updated at 04:09 PM ---------- Previous update was at 04:01 PM ----------

So I read that page:

0x100000, 0x200000, 0x400000, 0x800000, 0x1000000

And already set the fcs0 interface to 0x400000 from day 1 that I had these cards. Earlier I had the LPE10000 cards @ 2GB/s and they showed the same error.

Was looking at this page here: chaosisme.com blog

to see if there aren't any settables I could try but where as it showed me the /pci listing earlier, this time my power5 doesn't show /pci devices via this firmware. So I'm wondering if this isn't PCI related vs card related.

Cheers,
DH

Yes, but 0x400000 is not the largest possible buffer size to set. How about increasing it?

I'm compiling another targeting system. Blew away the old one. I need to get more logging at the time the Emulex / IBM system throws these and trying to eliminate some software as possible culprits.

After each failure, the LUN presented to the initiator is inaccessible until I restart SCST which in turn pulls down the QLogic drivers then brings them up again. This particular event is hinting that it isn't the IBM system or the Emulex cards, it's likely the target software, QLogic card or the PCIe bus.

I will be increasing the buffer to test with as soon as I'm done.

Cheers,
DH