[c-nsp] 7600/SRB4 not sending bgp updates

[Andrei Radu]

Hello everybody,

I've been have a ton of trouble with one of our 7600es (everyone who
hasn't please stand up :P) used as a border/exchange router. So
basically this box does OSPF/LDP and iBGP towards the rest of the
network and eBGP to a number of external peers.

The problem is that for new neighbors configured in a certain
peer-group the router does not send any updates, although it shows
prefixes as being advertised in the output of the "show ip bgp
neighbor A.B.C.D advertised-routes" command. In the output of "show ip
bgp neighbor A.B.C.D" the update counter remains at a steady 0:

  Message statistics, state Established:
    InQ depth is 0
    OutQ depth is 0
                         Sent       Rcvd
    Opens:                  1          1
    Notifications:          0          0
    Updates:                0         19
    Keepalives:            30        37
    Route Refresh:        0          0
    Total:                    31         57

and the total prefixes remains also at 0:

                                 Sent       Rcvd
  Prefix activity:               ----       ----
    Prefixes Current:             504          2 (Consumes 104 bytes)
    Prefixes Total:                 0          2
    Implicit Withdraw:            0          0
    Explicit Withdraw:            0          0
    Used as bestpath:         n/a          2
    Used as multipath:         n/a          0

I was able to reproduce this behavior using a bgp session with a test
router, and also found a workaround: configuring new peers in a
different peer-group with only the name and the name of the output
route-map changed, this forced the router to place the new neighbors
in a different update-group. Everything seemed to work ok until
earlier today when the workaround peer-group started behaving
identically :( .

I am sure it is a SRB4 bug and it looks kind of similar to the bgp
ghost bug seen on the SXH train, my bigger problem is I cannot open a
TAC case for this box as it is out of smartnet support.

Was anyone else hit by this ?  Does anyone have a more stable
workaround (short of reloading the box :) ) ? The only thing I could
find in the bug toolkit was CSCsm57494 which also seems similar but
does not contain too many details.

Best regards,
-- 
Andrei

"2+2=5, for extremely large values of 2 !"