Comware

Hi, guys.

I own a 5900AF switch:

5900AF-48XG-4QSFP+ Software: 2432-P06

I'm trying to set up, tcp mss for the vlan interfaces in the following way:

[core]int vlan2
[core-Vlan-interface2]tcp mss 1400
[core-Vlan-interface2]dis thi
#
interface Vlan-interface2
description UPLINK/WAN
ip address 19.21.15.9 255.255.255.248
tcp mss 1400
#
return

The problem is, that even doing it this way, the internet traffic in this VLAN, still uses the default mss of 1460 for all inbound/outbound traffic, can someone enlight me, why is this ?

Could this be a bug ? Because I believe, I've set it up properly ? Or did I miss anything ?

Hello @F1985P !

It seems you have set it up properly, but I am afraid in these switches this command only works when traffic goes through the CPU of a switch, e.g. for traffic destined to the switch or initiated from the switch itself (SSH, Telnet, BGP etc), so transit traffic won't be affected by this setting. The only exception is traffic over GRE tunnels where switch is an end-point of the tunnel and thus TCP MSS adjustment option affects that traffic too.

Our routers can adjust TCP MSS of transit traffic as per your expectations when traffic goes through a routed (L3) port, but in switches this command is limited to the over-the-CPU traffic.

I understand, but why doing it per interface, instead of doing it globally then ?

For example, in Huawei switches (it uses a similar CLI to HPE/H3C) you can set tcp mss globally (for cpu-initiated traffic). And then you have the interface adjustments for transit traffic.

The manual doesn't mention this anywhere, it simply says if you want to adjust the inbound/outbound parameter mss, you must use the desired command.

That is very true, the manual doesn't explicitly state this, I found this limitation in our case database. However, if you need to be 100% sure or if you need a documentary confirmation for purpose of network design, I advise you to open a case with our Support. Our Support engineers can send it to product development team and get official statement on this matter.

Do you know, if this apply to all switches without exception ? And if yes, this means the only way would be HP HSR/MSR line ?

According the information I have it is applicable to 5130-EI/HI, 5510-HI, 5900/5920. Not sure about 12900 though, but in chassis-based switches capabilities depend not only on MPU, but on capabilities of each LPU as well, so things may get complicated there depending which line cards are installed. Also I can't confirm it for newest 59xx series switches like 5945 and others. Again, such questions needs to be addressed using official process through Support department in order to get the final confirmation.