[c-nsp] N7k: Zero SVI counters

[Bernhard Schmidt]

Hey,

we're currently labbing Nexus 7k as replacement for Sup720 in our campus
enviroment and have hit a weird issue. In the lab the SVI counters
looked good enough, but now we have moved it into limited production.

Three 10GE interfaces, all of them .1q trunks, two have dedicated OSPF
vlans to neighbors, the third one has yet another VLAN for users. The
users in this network can work fine, but the "show interface vlan xxx"
counters only show one direction. That is it shows only input packets on
the user SVI

  Load-Interval #2: 5 minute (300 seconds)
    input rate 38.60 Kbps, 68 pps; output rate 0 bps, 0 pps
  L3 Switched:
    input: 839253 pkts, 254101754 bytes - output: 0 pkts, 0 bytes
  L3 in Switched:
    ucast: 833112 pkts, 252581673 bytes - mcast: 6141 pkts, 1520081
bytes
  L3 out Switched:
    ucast: 0 pkts, 0 bytes - mcast: 0 pkts, 0 bytes

and only output packets on the OSPF vlans

  Load-Interval #2: 5 minute (300 seconds)
    input rate 0 bps, 0 pps; output rate 11.83 Kbps, 20 pps
  L3 Switched:
    input: 0 pkts, 0 bytes - output: 785296 pkts, 245273374 bytes
  L3 in Switched:
    ucast: 0 pkts, 0 bytes - mcast: 0 pkts, 0 bytes
  L3 out Switched:
    ucast: 785296 pkts, 245273374 bytes - mcast: 0 pkts, 0 bytes

at the moment the box is not having any transit traffic so I can't tell
whether we will see that, but it looks weird enough. It should see at
least the occasional OSPF hello. We have of course verified that the
traffic is going through this particular router etc.

The user network is dualstacked, MTU 1500, DHCP relay, IPv4 and IPv6
netflow and uRPF. The OSPF networks are dedicated IPv4 and IPv6, MTU
9216, IPv4 and IPv6 netflow. We're running NX-OS 5.2(1), the 10GE ports
are on N7K-M108X2-12L.

The SNMP counters are zeroed too, but Netflow records look fine.

Our Cisco SE is searching, but maybe someone else has already seen that.

Bernhard