As I understood it, VPLS Control Word would fragment an Ethernet packet, then reassemble at the far end of the VPLS tunnel.
Without changing any L2MTU settings, when I set ping to do not fragment, I can only get 1472 across. I understand the overhead that VPLS has, but I thought using the Control Word would resolve that issue for me.
I will increase the L2MTU to the capabilities of the gear between, but I want to work on one “issue” at a time.
Yes, with control word enabled packets can be fragmented and reassembled.
Make sure that you have set MPLS MTU on all interfaces in LS path.
It was set to 1508. I increased it to 1518, which I understand is the MTU of Ubiquiti AirOS 5.3 gear.
I increased the L2MTU to 1518 on all interfaces I could. For some reason, on my 493 and 750, I couldn’t change the L2MTU, but it was higher than 1518.
I did try my test again, but it still wouldn’t let me pass a ping larger than 1472 without fragmenting. I suspect this is because that’s all regular Ethernet allows. I’ll have to find a way to test higher than that.
As far as I know ubiquity does not allow to pass frames larger than 1500 if not explicitly allowed.
Decrease all MPLS MTU to 1500 for all interfaces (in ‘/mpls interface’ menu).
I’ll ask them if I have to do anything on AirOS 5.3 to allow more than 1500. They have told me the MTU is 1518 on 5.3.
Newer hardware with AirOS 5.4 or 5.5 allows over 2000. They made it a point to email me when that hardware change was made.
I have attached a screen shot showing the MTU settings on an AirOS 5.3.5 device. It will accept an MTU of between 64 and 1518. Is 1518 what I would set my L2MTU and MPLS MTU to?
With L2MTU and MPLS MTUs set to 1518, it looks like I can traffic an MTU of 1486 (using pings set to not fragment).
I set my Windows MTU to 1518 and pinged through the MPLS\VPLS cloud.
Does this sound right?
What happened to control word auto fragmenting and reassembling?