S6700EI ping the loopback ip address failed with the big packets in the mpls situation

Publication Date:  2016-04-25 Views:  560 Downloads:  0
Issue Description

S6700EI is PE or P of the IP/MPLS backbone network. If the customer tried to ping the loopback IP address with packet’s size 7000byte, the last packet always dropped .And at the same time there is cpu-defend drop statistic under mpls-fib-hit. However the customer tried to ping the interface IP address with packet’s size 7000byte, it is always successful and no drop statistic under mpls-fib-hit.

(1)ping the loopback IP address:

[Switch]ping -f -s 7000 1.1.1.1                                           

  PING 1.1.1.1: 7000  data bytes, press CTRL_C to break                         

    Reply from 1.1.1.1: bytes=7000 Sequence=1 ttl=255 time=14 ms               

    Reply from 1.1.1.1: bytes=7000 Sequence=2 ttl=255 time=11 ms               

    Reply from 1.1.1.1: bytes=7000 Sequence=3 ttl=255 time=14 ms                

    Reply from 1.1.1.1: bytes=7000 Sequence=4 ttl=255 time=27 ms               

    Request time out                                                           

                                                                               

  --- 1.1.1.1 ping statistics ---                                              

    5 packet(s) transmitted                                                    

    4 packet(s) received                                                       

    20.00% packet loss                                                         

round-trip min/avg/max = 11/16/27 ms                                       

(2)ping the interface IP address:

[Switch]ping -f -s 7000 100.0.0.2                                         

  PING 100.0.0.2: 7000  data bytes, press CTRL_C to break                      

    Reply from 100.0.0.2: bytes=7000 Sequence=1 ttl=254 time=2 ms              

    Reply from 100.0.0.2: bytes=7000 Sequence=2 ttl=254 time=2 ms              

    Reply from 100.0.0.2: bytes=7000 Sequence=3 ttl=254 time=2 ms              

    Reply from 100.0.0.2: bytes=7000 Sequence=4 ttl=254 time=1 ms              

    Reply from 100.0.0.2: bytes=7000 Sequence=5 ttl=254 time=2 ms              

                                                                                

  --- 100.0.0.2 ping statistics ---                                            

    5 packet(s) transmitted                                                    

    5 packet(s) received                                                        

    0.00% packet loss                                                          

    round-trip min/avg/max = 1/1/2 ms 

 

The software version of S6700EI is V200R005.

This is the topology:


Alarm Information

None

Handling Process

(1)    We checked the cpu-defend statistic and found that when the customer ping the loopback address it matched the packet type of mpls-fib-hit, and there are drop statistics on the packet type of mpls-fib-hit. However, when the customer ping the interface address it matched the packet type of ICMP, and there are no drop statistics on the packet type of ICMP.

(1)the cpu-defend statistic when ping the loopback IP address:

(2)the cpu-defend statistic when ping the interface IP address:

  

(2)    Why the customer ping different IP address, the ICMP packets matched different packet type. Because the switch only distributes mpls label to the IP host routes with 32-bit addresses by default, and won’t distributes mpls label to the direct route.

The mpls lsp:

The mpls configuration:

 

(3)    The packet will include the mpls head when you ping the loopback address, but the packet won’t include the mpls head when you ping the interface address. The priority of mpls-fib-hit is higher than ICMP, so the packets with the mpls head are matched the packet type of mpls-fib-hit.

 

(4)    Why the last packet always dropped when the customer ping the loopback IP address. Because the cpcar of the packet type of ICMP is larger than mpls-fib-hit.

(1)the cpcar of ICMP:

(2) the cpcar of mpls-fib-hit:

 

(5)    For huawei switch, the data packets are transferred by the chips, and it won’t affect the business and the ping test is only used to check the network is connected well.

Root Cause

The switch only distributes mpls label to the IP host routes with 32-bit addresses by default, and won’t distributes mpls label to the direct route. The ICMP packets with the mpls head are matched the packet type of mpls-fib-hit, but the ICMP packets without the mpls head are matched the packet type of ICMP. And the cpcar of the packet type of ICMP is larger than mpls-fib-hit.

Solution

This is a normal behavior for our switch and it won’t affect the business.

Suggestions

None.

END