Upgrading rp3440 with DS2300

[nibble]

Hi everyone,

In relation to this thread that I opened few days back: http://forums12.itrc.hp.com/service/forums/questionanswer.do?admit=109447627+1203634236794+28353475&threadId=1203474

Unfortunately, I still don't have the solution to remove the scsi bus errors. based on history of these systems, cables, terminators, BCCs and scsi channels have been changed already but still to no avail.

There are 2 servers of exactly the same configuration and the same scsi bus errors.
Development
Server: RP3440 (A7137A), UX11.23, 1000Mz, 2CPUs, 8GbRAM, 2x73G U320 (internal)
Storage: DS2300, 14 x 73Gb U320 15K, Split bus mode, BCCs connected to 2 x Ultra160 LVD A6828-60101

Production
Server: RP3440 (A7137A), UX11.23, 1000Mz, 2CPUs, 8GbRAM, 2x73G U320 (internal)
Storage: DS2300, 14 x 73Gb U320 15K, Split bus mode, BCCs connected to 1 x SCSI C1010 Ultra160 Wide LVD A6829-60101 (dual channel)

Both are running SAP application. For now, my investigation leads that the cause of the scsi reset is due to high IO activity.
Im planning to recommend for replacement of storage and use of fiber instead of scsi. any ideas on how to do this? Any response for the troubleshooting for my previous thread or new design recommendation is highly appreciated.

[TTr]

Did you consider to upgrade the dual port U160 controller by two PCI U320 controllers? (single or dual port). Then you can put each half of the DS2300 on a separate PCI slot. The rp3440 has lots of high speed I/O slots.

Also is it possible that on the dev server you limit the I/O down to only one mirrored pair of drives and exersize the I/O to reproduce the errors. Then gradually allow two and three mirrored pairs in the mix. You should get a feel about when the errors start to occur and where the I/O is.

Is the I/O uniform across all 7 pairs or is it concentrated in one or two pairs?

Have you checked the SCSI queue length in glance and the I/O per disk pair. (hit the "u" key in the main glance screen)

[Duncan Edmonstone]

There's no point going for a U320 SCSI card with the DS2300, as it is only U160 capable anyway.

The fact that you see the problem on both systems suggests config issues to me rather than anything else...

1) Has the firmwars on the BCCs been flashed to the latest version? (I guess yes given all the other stuff you have done)

2) How do you have the dip switches set on the BCCs? Do they match? See the attached from the manual

HTH

Duncan