[c-nsp] netflow issues on WS-F6700-DFC3CXL - 7600

[Ben Steele]

I have already lodged a TAC for this (actually on my second TAC for same
issue) but I thought i'd throw out here to see if anyone else has seen this
as it has me perplexed at the moment.

 

Problem: Netflow collector stops receiving flows from DFC on 7609-S but
continues to receive flows from RSP, identical router with same hardware and
code has no problems exporting netflow via DFC and RSP, all is well on that
chassis.

 

Have tried software upgrade from SRB2 to SRB3, problem still existed, moved
10Gb int onto non DFC line card to let RSP process netflow and it was no
problems, pump out netflow all day long, move onto DFC and you get netflow
for about 7-10 hours (well it was done at very early hours each time and
then would die as the traffic built up in the morning) then it stops
exporting flows for the DFC only.

 

Weird thing is a sh mls netflow ip mod 1 (module where dfc is) is full of
flows, and the table-contention info is showing it as creating netflows and
not having failures, TCAM utilization is nice and low at around 7-10%, I did
change mls aging timers to get this but that had no affect on netflow, it
was more because I was hitting TCAM limits on the RSP.

 

When the DFC failed exporting again this morning(around 10am) after I
powered down the line card and brought it back up at 1am I checked the pps
going out the dedicated netflow collector interface, I then turned off "ip
flow ingress" on the DFC interface and didn't see a change in that interface
output which is leading me to believe that it is indeed not making it out of
the router despite the router thinking all is well.

 

So as mentioned software upgrade has occurred, also an entire new line card
was sent out via RMA from TAC (WS-X6704-10GE + DFC) and replaced and we
still have the same problem, yet I don't have the problem on an exact same
model and basically same config sitting next to it.

 

Idprom shows the hardware revision to be different on the DFC's between the
2 chassis but the new RMA card was a different revision again and still have
the same issue so...?

 

Any ideas? J

 

Cheers

 

Ben