Monitor-session over pseudowire causing LACP to fail on remote port

[7tsommer]

I'm trying to configure a monitor-session over a pseudowire between an ASR9K and a NCS5508, both running IOS-XR 24.1.1. 

On the ASR, the monitor source is a 100G bundle and the destination is a l2vpn xconnect to the NCS

On the NCS, the xconnect from the ASR terminates on a dot1q tagged subinterface on a 100G bundle facing our switch where our Wireshark lives.

When I activate the pseudowire, the bundle on the NCS immediately starts bouncing LACP. I thought maybe the mirrored traffic was overwhelming the link, but the source bundle on the ASR is only carrying a measly 8kbps in and 14kbps out. There are no other sources in the monitor-session. 

Here are the details:

ASR9K config (note that I've tried with and without the pw-class in the l2vpn config, both with the same results):

interface Bundle-Ether103
 mtu 9216
 ipv4 address 10.253.101.252 255.255.255.254
 ipv6 nd ns-interval 9000
 ipv6 nd reachable-time 2700000
 ipv6 address fd00::10:253:101:252/127
 ipv6 enable
 lacp period short
 lldp
  enable
 !
 monitor-session TEST ethernet
!
monitor-session TEST ethernet
 destination pseudowire
!
l2vpn
 pw-class SPAN
  encapsulation mpls
   transport-mode ethernet
  !
 !
 xconnect group SPAN_TEST
  p2p SPAN_TEST
   monitor-session TEST
   neighbor ipv4 10.251.0.11 pw-id 9999
    pw-class SPAN
 !

Some show commands from the ASR

Showing only a single source for the monitor session (currently down since remote end is shutdown):

RP/0/RSP0/CPU0:LAB-ASR01#show monitor-session TEST status
Thu Jun 19 13:46:57.679 EDT
Monitor-session TEST
Destination pseudowire  (down)
Session has not been set up in the hardware
================================================================================
Source Interface      Dir   Status
--------------------- ----  ----------------------------------------------------
BE103                 Both  Not operational (destination not active)
RP/0/RSP0/CPU0:LAB-ASR01#

Showing small amount of packets transmitted (not sure if the last line is there because the session is currently down):

RP/0/RSP0/CPU0:LAB-ASR01#show monitor-session TEST count      
Thu Jun 19 13:47:02.104 EDT
Monitor-session TEST
  Bundle-Ether103
    Rx replicated: 1493 packets, 191699 octets
    Tx replicated: 1320 packets, 431132 octets
    Non-replicated: 0 packets, 0 octets
Show request failed. Error: 'SPAN' detected the 'fatal' condition 'Invalid capabilties received from the platform'

NCS config:

interface Bundle-Ether201
 description OLT1-399-LAB LAG
 mtu 9216
 lacp period short
 lldp
  transmit disable
 !
 flow datalinkframesection monitor SFLOW_MONITOR sampler FLOW_SAMPLER-1-4000 ingress
!
interface Bundle-Ether201.9999 l2transport
 encapsulation dot1q 999
!
l2vpn
 xconnect group SPAN_TEST
  p2p SPAN_TEST
   interface Bundle-Ether201.9999
   neighbor ipv4 10.251.0.8 pw-id 9999
   !
  !
 !
!

Show commands from the NCS

Showing amount of packets output:

Bundle-Ether201.9999 is administratively down, line protocol is administratively down 
  Interface state transitions: 262
  Hardware is VLAN sub-interface(s), address is 0caf.31b6.4c82
  Layer 2 Transport Mode
  MTU 9220 bytes, BW 100000000 Kbit (Max: 100000000 Kbit)
     reliability Unknown, txload Unknown, rxload Unknown
  Encapsulation 802.1Q Virtual LAN,
    Outer Match: Dot1Q VLAN 999
    Ethertype Any, MAC Match src any, dest any
  loopback not set,
  Last link flapped 00:12:49
  Last input never, output 00:12:27
  Last clearing of "show interface" counters 00:42:20
     0 packets input, 0 bytes
     0 input drops, 0 queue drops, 0 input errors
     977224 packets output, 4874427646 bytes
     0 output drops, 0 queue drops, 0 output errors

Console messages occurring as soon as Be201.9999 is no-shut:

RP/0/RP0/CPU0:LAB-NCS01(config-subif)#no shut
RP/0/RP0/CPU0:LAB-NCS01(config-subif)#RP/0/RP0/CPU0:Jun 19 
RP/0/RP0/CPU0:LAB-NCS01(config-subif)#commit
RP/0/RP0/CPU0:Jun 19 17:26:57.184 UTC: BM-DISTRIB[1186]: %L2-BM-6-ACTIVE : HundredGigE0/1/0/1 is no longer Active as part of Bundle-Ether201 (Partner is not Synchronized (Waiting, Standby, or LAG ID mismatch)) 
RP/0/RP0/CPU0:Jun 19 17:26:59.246 UTC: BM-DISTRIB[1186]: %L2-BM-6-ACTIVE : HundredGigE0/1/0/1 is Active as part of Bundle-Ether201 
RP/0/RP0/CPU0:Jun 19 17:27:01.173 UTC: BM-DISTRIB[1186]: %L2-BM-6-ACTIVE : HundredGigE0/1/0/1 is no longer Active as part of Bundle-Ether201 (Partner is not Synchronized (Waiting, Standby, or LAG ID mismatch)) 
RP/0/RP0/CPU0:Jun 19 17:27:03.220 UTC: BM-DISTRIB[1186]: %L2-BM-6-ACTIVE : HundredGigE0/1/0/1 is Active as part of Bundle-Ether201 
RP/0/RP0/CPU0:Jun 19 17:27:05.180 UTC: BM-DISTRIB[1186]: %L2-BM-6-ACTIVE : HundredGigE0/1/0/1 is no longer Active as part of Bundle-Ether201 (Partner is not Synchronized (Waiting, Standby, or LAG ID mismatch)) 
RP/0/RP0/CPU0:Jun 19 17:27:07.225 UTC: BM-DISTRIB[1186]: %L2-BM-6-ACTIVE : HundredGigE0/1/0/1 is Active as part of Bundle-Ether201 
RP/0/RP0/CPU0:Jun 19 17:27:09.213 UTC: BM-DISTRIB[1186]: %L2-BM-6-ACTIVE : HundredGigE0/1/0/1 is no longer Active as part of Bundle-Ether201 (Partner is not Synchronized (Waiting, Standby, or LAG ID mismatch)) 

Any ideas how to troubleshoot or fix this? 

[7tsommer · Accepted solution]

I figured out the problem. The source port being monitored is a LACP bundle with no subinterfaces. Mirroring the full bundle and sending it across the psuedowire and out a subinterface on another bundle caused that destination bundle to get confused due to the mirrored LACP protocol frames. When I changed the source to a subinterface, the destination bundle no longer bounced. 

This solved the immediate problem, but we would still like to be able to monitor whole bundles. To solve that, we are going to make a dedicated physical, non-LACP connection between our NCS and the switch where our Wireshark lives so we no longer have to make our monitor destination a vlan subinterface on an existing bundle.