Solaris

I've got lots of scsi transport errors on the E2900 - the HBA was replaced once out of curiosity but the problem still exists.

Feb 15 07:03:11 ivsdbch2 scsi: [ID 243001 kern.warning] WARNING: /ssm@0,0/pci@18
,600000/SUNW,jfca@1,1/fp@0,0/ssd@w5005076300c889cb,4 (ssd21):
Feb 15 07:03:11 ivsdbch2 SCSI transport failed: reason 'tran_err': retryi
ng command
Feb 15 09:45:34 ivsdbch2 scsi: [ID 243001 kern.warning] WARNING: /ssm@0,0/pci@18
,600000/SUNW,jfca@1,1/fp@0,0/ssd@w5005076300c489cb,5 (ssd26):
Feb 15 09:45:34 ivsdbch2 SCSI transport failed: reason 'timeout': retryin
g command


I suspect this is the problem:

Of the two HBAs, these are dual-port - one of them is plugged into the 33Mhz pci while the other is in the 66Mhz, as shown here from prtdiag.


0 pci 66 PCI5 SUNW,jfca (scsi) FCX2-6562-L
0 pci 66 PCI5 SUNW,jfca (scsi) FCX2-6562-L
0 pci 33 PCI4 SUNW,jfca (scsi) FCX2-6562-L
0 pci 33 PCI4 SUNW,jfca (scsi) FCX2-6562-L


There isnt any other available 66Mhz slots if I want to move PCI4 over, so my only option is to move PCI5 to PCI3, which is a 33Mhz slot. I wonder if moving down to the 33Mhz would degrade any performance.

Anyone know if it's okay to plug these HBAs to the 33Mhz pci slots?

Thanks.

Given those error messages, and the fact that you've tried replacing the controller, I would try:

Replace the GBIC on the disk side.

Replace ssd21. It might be failing in such a way as to screw with other disks' use of the bus. (When you get bursts of error messages, does it always start with ssd21?)

(I've never used an E2900, my experience is with earlier Suns. But when I've had bursts of error messages from multiple disks in one FC array, is was always one of those two things.)

Also - googling for that JNI board shows that 66 Mhz PCI is considered legacy for it, so you really don't want it in an even slower slot.