cancel
Showing results for 
Search instead for 
Did you mean: 
cancel

Why do I see message "L2 MTU is near HW limit" and how to fix it.

934
Views
0
Helpful
0
Comments

Preface

Since introducing DDTS CSCus57232 XR software reporting following message:

 

LC/0/0/CPU0:Oct 31 15:54:03.752 GMT: vic_1[372]: %L2-ETHERNET-6-MTU_NEAR_HW_LIMIT : The MTU of 9220 on TenGigE0_0_1_3 is over the H/W limit, and hence large frames may be dropped as oversized frames

 

Typical scenario

 

Let's consider scenario when this message can appear and how to fix it.

 

Here is the typical configuration which leads to this message.

 

RP/0/RSP0/CPU0:adum#sh run int GigabitEthernet0/1/1/9*
Fri Dec  2 16:04:06.728 CET
interface GigabitEthernet0/1/1/9
 cdp
 mtu 9216
 ipv4 address 63.1.1.2 255.255.255.0
!
interface GigabitEthernet0/1/1/9.111
 description ---ankomaro HWLIMITTEST
 ipv4 address 10.10.10.1 255.255.255.252
 encapsulation dot1q 111
!
interface GigabitEthernet0/1/1/9.200 l2transport
 encapsulation dot1ad 200 dot1q any

 

Here is the parent interface with MTU 9216 configured and subinterface (which frequently can be configured later and operator just missed the fact of MTU configuration on the parent interface).

 

Let's send some pings through parent and sub interfaces with df-bit set and not:

 

RP/0/RSP0/CPU0:adum#ping 63.1.1.1 size 9202 df-bit
Fri Dec  2 16:03:08.743 CET
Type escape sequence to abort.
Sending 5, 9202-byte ICMP Echos to 63.1.1.1, timeout is 2 seconds:
!!!!!
Success rate is 100 percent (5/5), round-trip min/avg/max = 2/2/3 ms

RP/0/RSP0/CPU0:adum#ping 10.10.10.2 size 9202 df-bit
Fri Dec  2 16:03:23.361 CET
Type escape sequence to abort.
Sending 5, 9202-byte ICMP Echos to 10.10.10.2, timeout is 2 seconds:
.....
Success rate is 0 percent (0/5)

 

RP/0/RSP0/CPU0:adum#ping 10.10.10.2 size 9202
Fri Dec  2 16:03:41.610 CET
Type escape sequence to abort.
Sending 5, 9202-byte ICMP Echos to 10.10.10.2, timeout is 2 seconds:
.....
Success rate is 0 percent (0/5)

 

RP/0/RSP0/CPU0:adum#ping 10.10.10.2 size 9198
Fri Dec  2 16:03:56.364 CET
Type escape sequence to abort.
Sending 5, 9198-byte ICMP Echos to 10.10.10.2, timeout is 2 seconds:
!!!!!
Success rate is 100 percent (5/5), round-trip min/avg/max = 3/3/3 ms

 

As it can be observed router drops the oversized packets regardless of the df-bit until 4 bytes are subtracted and this causes the issue which can be hard to troubleshoot.

 

The Max MTU size is limited by the interface driver and can be derived from this command output:

 

 RP/0/RSP0/CPU0:adum#sh controllers GigabitEthernet0/1/1/9 control | i Max MTU
Fri Dec  2 16:05:09.085 CET
    Max MTU: 9216 bytes

 

What happens with MTU size reported on the interface after MTU configuration (Typhoon LC will show MTU + 4 FCS bytes):

 

RP/0/RSP0/CPU0:adum#sh controllers GigabitEthernet0/1/1/9 | i MTU
Fri Dec  2 16:05:15.070 CET
    MTU: 9216

RP/0/RSP0/CPU0:adum#sh int GigabitEthernet0/1/1/9.111 | i MTU
Fri Dec  2 16:06:43.522 CET
  MTU 9220 bytes, BW 1000000 Kbit (Max: 1000000 Kbit)

 

Here under subinterface you can see MTU 9220 = 9216 + 4 (vlan tag) and hense frame size of 9216 + VLAN tag (4 bytes) will be dropped as driver has no room space for it.

 

How to fix it?

 

On the parent interface lower MTU by the number of bytes needed for the vlan (depends on how many tags will be in the frame).

 

After reconfiguring MTU on the parent interface to 9212:

 

RP/0/RSP0/CPU0:adum#sh int g0/1/1/9 | i MTU
Fri Dec  2 16:10:31.904 CET
  MTU 9212 bytes, BW 1000000 Kbit (Max: 1000000 Kbit)
RP/0/RSP0/CPU0:adum#sh int g0/1/1/9.111 | i MTU
Fri Dec  2 16:10:47.864 CET
  MTU 9216 bytes, BW 1000000 Kbit (Max: 1000000 Kbit)

 

Pings with 9216 frame size run smoothly:


RP/0/RSP0/CPU0:adum#ping 10.10.10.2 size 9202
Fri Dec  2 16:09:47.464 CET
Type escape sequence to abort.
Sending 5, 9202-byte ICMP Echos to 10.10.10.2, timeout is 2 seconds:
!!!!!
Success rate is 100 percent (5/5), round-trip min/avg/max = 2/3/6 ms