cisco write accelaration generates scsi-retry errors

[vijayabalan]

After implementing write accelaration between mds9509 on DR sites (synchronous SRDF) , we are seeing a lot of scsi-retry errors. Can you see a reason, why?

[dmcloon]

There are a few gotchas out there for FCIP WA, but it depends on which SAN-OS release your running.

[oerkan]

Hi,

Unfortunately, It is difficult to predict up what is causing the SCSI errors without doing further investigation. I would suggest you to open a problem ticket with the TAC and provide the following information from the time of the problem.

- terminal session-timeout 0

- clear counters interface fcip for all fcip interfaces)

- show interface fcip counters

- attach module

- terminal session-timeout 0

- debug ips fcip write-acc-err port 1

- debug ips fcip write-acc-err port 2

- show tech-support

- show interface fcip counters (for all fcip interfaces)

- show ips stats tcp interface gigabitethernet X/Y detail (for all gige

interfaces)

- show ips stats dma-bridge interface gigabitethernet X/Y (for all gige

interfaces)

Thanks and regards,

Hakan.

[vijayabalan]

This is a direct fiber dwdm connection between R1 (primary) and R2 (secondary) site. Does that make a difference?

[Arueda]

i do not know what code you are running on sym, but there are a few problems with different code releases and not all of them like WA.

a couple of things to keep in mind when you go over the wan.

check latency, buffer to buffer credit usage, with WA enabled and with WA disabled and see if they are similar. many times i have found that the problem is the "queuing" at the switch what makes scsi to retry.

also what are the commands that are being retried?

if you are using windows it can be that your host is sending rsns and for some reason they are being sent over the wire to your remote site although this is a very rare situation if it is happening ic will cause a lot of issues.

if you have a Cisco port analyzer get a trace and read it to see if you can see specific conversations that are the problem.

also the problem could be at the physical layer have you checked this?

if you want and can i could check you trace to see if i cn see anything familiar.

i do not use syms. but i do use CLARiiON and they share quite a bit of code nowadays

[affox]

What distance between Symm bixis and what technology between MDS boxis (FC, FC over C(D)WDM, FCIP over ?).

What delay You have between MDS boxis?