ASR 1006 reodering traffic with fragments

[Eugene Golovachev]

We have an issue with ASR1006 reordering fragmented UDP multicast. 

   Source --> Nexus --> (Ten0/0/0) ASR (Gig0/1/6) --> Receiver

UDP is a multicast traffic consists from relative small (~300 byte) and huge fragmented (~2000 byte) packets. MTU 1500.

Dump from outside Nexus port shows that packet order is correct.

Sometimes from outside ASR interface we have situation when smaller packets goes first over large and order is incorrect. Not very often, but several times every minute. Application is sensitive to order. 

Example:

  packet seq 1 

  packet seq 2 fragment start (1500 byte)

  packet seq 3 (300 byte)

  packet seq 2 fragment end (700 byte)

  packet seq 4 

after reassembly seq: 1, 3, 2, 4

Router is not loaded: interface  rate 36000 bits/sec, 9 packets/sec. Queueing strategy: fifo or class based wfq - the same

interface TenGigabitEthernet0/2/0
 ip address 192.168.0.1 255.255.255.252
 ip nat inside
 ip flow ingress
 ip pim sparse-mode
 ip ospf cost 1
 cdp enable

!
interface GigabitEthernet0/1/6
 description test server
 ip address 10.0.0.1 255.255.255.0
 ip pim passive
 load-interval 30
 negotiation auto

Does anyone know a way to troubleshoot and correct this behavior? I have no clue.

[Eugene Golovachev]

I found the cause of reordering - NAT, even nat rules are not for multicast and there's no active translations.

After disabling NAT on ingress interface, correct order of packet flow was restored.

Bug or feature?

[ivanher]

I have a similar case, but in my case I do not have NAT but vrf and seems as per packet captures same out-of order is happening but seems expected, thing is that it is not happening with other ASR and it leads me to thing that it depends on the network conditions.

in your case nat was the problem.