05-12-2010 05:37 PM
we are seeing intermittent scsi errors on hosts that use the ISL to get to the target. If the initiator does not have to go over the ISL we do not see these scsi errors.
what do i need to look for in troubleshooting this ?
05-14-2010 02:14 AM
Just a starting point:
Are you using Trunking on your ISLs or Exchanged-based routing, Port-based routing?
Have you set In Order Delivery? type "iodshow" to see.
What is the FOS version you are using on your switches? Check if you have compatible FOS versions on your switches.
If you're in interop mode with McData switches that may also have some conditions to check.
If you're getting SCSI errors which is a high-level layer on the FC stack then I would bet more to some host-side issue. So take a look at your multi-pathing software, patches, OS issues, etc.
Check your compatibility list between disk array-HBA-Server OS-application-Volume Manager, etc. This are typical causes of these kind of problems more than the switch itself.
05-14-2010 08:41 PM
iodshow is not set. we are using trunking on the ISL. FOS version used is 6.2.1b. Its a brocade native fabric.
we are using trunking on the ISL.
We did see a bunch of ports (around 10) with high error rates and are fixing those. Since then the scsi error rates have died down.
Question is: Could a single port or a handleful of ports with issues cause fabricwide issues and timeouts for host?
05-15-2010 08:00 AM
This kind of condition could cause the flood of a lot of RSCNs and therefore can potentially generate a fabric wide issue. So yes! Take a look at the cables and also at the SFPs that may have gone bad.
Good luck with the issue and thanks for posting how did you address the problem.