12-05-2014 08:05 PM
So I have had a case open with TAC for months now on this. Whenever I try to rate limit in the very next piece of gear directly downstream of the MLX, the pings start getting discarded and never get real high. If I rate limit in the MLX and not anywhere downstream, it behaves as expected. If I plug a computer directly into the MLXe-4 it also behaves as expected. Originally the thought was it was a compatibility issue with Occam, Adtran or Cisco but they sent me a Brocade FCXS switch and it behaved identical to the others. It happens with both the 1 and 10 gig ports on all cards. All cards have been replaced including SFM and management cards. This has been going on since day 1 and we have been through 4 software releases. We have X cards.
I just can't imagine the backplane is bad as been suggested. I can duplicate issue very easily by raising traffic levels to around 50% or more of the downstream limiter. Now I am hearing from them that this is normal behavior. Wireshark shows the packets make it out of the MLX and get discarded before entering the downstream device.
Does anybody have any ideas? I am desperate as customers notice and complain. No issues like this when we had a Brocade CER. Ping times would actually raise to 1000-2000 ms but very seldom drop unless there was a physical line issue.
I should also mention while loss is happening we can ping from test computers to a VE (on MLX) and back with no loss. So the loss is apparent when pinging through the MLX CPU, although as Wireshark has told us many times, the MLX itself is not actually dropping packets.