[A-series] LACP over VPLS?

[MichaelM55]

Hello,

is LACP over multiple VPLS instances supported on Comware?

e.g. imagine a IRF stack of two switches, which you would like to connect with each other (see picture)

[paulgear]

I don't have an answer for you, but kudos for an interesting question. :-)

Do the VPLS circuits appear as layer 2 links? If so, i would definitely try and see how it goes.

[manuel.bitzi]

In my opinion and my knowledge of VPLS, it should work. But I had never tried it out.

br

Manuel

[Peter_Debruyne]

Hi,

LACP uses the ethernet link-local range 01:80:C2 (just like 802.1x, lldp, stp etc), so most intelligent switches will drop packets from this range.

When you are using comware devices for the VPLS setup, (or a QinQ setup would be similar), you can configure the PE devices with bpdu-tunneling.

When the CE device sends an LACP packet, the PE device with bpdu-tunnel feature will re-write the dstmac to a custom dstmac, so it can safely pass the backbone. The remote PE device will again re-write the dstmac to the original protocol dstmac (01:80:C2:xx:xx:xx), so the remote device on the VPLS link would not see the difference.

You may want to configure the IRF devices with the "lacp period short" command, to detect any cloud failures faster than the default 90sec lacp timeout.

Other suggestion would be to enable OAM on the interfaces, which is some kind of BFD (hello exchange) between the 2 CE (the IRF switches in your picture) devices, which will be able to detect a VPLS failure in about 1 sec.

Best regards,Peter.

[Trinh_Nguyen]

Michael

Me too, you should read my post here:

http://h30499.www3.hp.com/t5/Comware-Based/Problem-with-STATIC-LACP-Comware-5-20-105-1205L01-US/td-p/5919929

In my experience, dynamic lag will not work because BPDU drop; static lag works, but if any of your VPLS not working  your switch still forwarding packets to that links and data lost will occur

TN