IOS-XRv xconnect packet loss

[kim]

Hi guys,

thanks for an awesome resource!

I've run into a small problem in my L2 lab. I've managed to find an XRv image with a working L2 dataplane but with packet loss it seems.

I was testing the XRv image with a multi segment xconnect setup. So i effectively stitch 2 PW's together on a 3rd PE. Anyway when pinging across the evc, there is connectivity. But packets are dropped at the ACs towards the CPEs. Either side shows this issue, two different PEs. This is what i see:

<snip>

CPE6#ping 1.1.1.7 rep 10000
Type escape sequence to abort.
Sending 10000, 100-byte ICMP Echos to 1.1.1.7, timeout is 2 seconds:
.!!!!!!!!!!!!!!!!!!.!!!!!!!!!!!!!!!!!!.!!!!!!!!!!!!!!!!!!.!!!!!!!!!!!!
!!!!!!.!!!!!!!!!!!!!!!!!!.!!!!!!!!!!!!!!!!!!.!!!!!!!!!!!!!!!!!!.!!!!!!
!!!!!!!!!!!!.!!!!!!!!!!!!!!!!!!.!!!!!!!!!!!!!!!!!!.!!!!!!!!!!!!!!!!!!.
!!!!!!!!!!!!!!!!!.!!!!!!!!!!!!!!!!!!.!!!!!!!!!!!!!!!!!!.!!!!!!!!!!!!!!
!!!!.!!!!!!!!!!!!!!!!!!.

<snip>

The lab runs in esxi 6.5 on a PowerEdge R630 with 2 CPUs and 256Gb memory. The host shows free cpu, mem and disk, plenty infact. PEs are XRv 6.4.1 while the CPEs are CSR1000Vs 16.06.04

Does anyone have any suggestions as to why this happens? Or what can be done? To me it looks like some kind of buffer issue. It looks pretty even and keeps going like that for the duration. I haven't seen any other issues yet.

Any input appreciated!

Thanks in advance.

Cheers

/Kim Tingkær

[Ashish Panda · Accepted solution]

Hi,

Could you add a back door link between the csrv CPEs? There use to be some data plane forwarding throttling with default license.
If you don’t see ping loss over the back door link between the CPEs then that will rule out the CPE issues.

In that case more troubleshooting will be needed to find out the router where the packet is dropped. You can find this info from the interface counters easily.

[kim]

Hi Ashish :)

You are absolutely right :) It must be the rate limit i'm running into. I've attached the wireshark dump showing the ping traffic hitting the limit.

Thanks.

Cheers

[Adam Vitkovsky]

Could even be the LPTS rate limitter do you see a similar behavior with
regular ping?

[smilstea]

LPTS only comes into play when the packets have to be punted up to the CPU (either destined to that router, fragmentation, etc), it shouldn't affect transit traffic between the two CPEs.

Sam

[kim]

Hey guys,

no problems experienced when pinging from PE to PE.

It most definitely is the license limitation :)

Thanks again!

/Kim