lacp not implemented

[linqingxuan]

I found that lacp is not implemented on the vpp platform. How to implement teamd with vpp

[yue-fred-gao]

LACP is tricky to support. SONIC implements LACP in control plane but vpp implements it in data plane. It doesn't punt LACP packets so SONIC will see LAG interface down because it is not receiving LACP packets. There is also complication involved the host path from vpp to kernel interface. @AkeelAli is investigating it.

[linqingxuan]

I have also investigated this lacp problem recently. Now I feel that sonic needs to be modified less. The vpp plug-in to implement the new bond mainly has two aspects: 1. Support sai control up and down, 2. lacp message is sent to the physically mapped tap. But I just started to try to write this plug-in recently.

[linqingxuan]

Hello @AkeelAli , I hope to help you in the lacp research and complete it together.

[AkeelAli]

Hi @linqingxuan, we had a code-a-thon last week during which I was able to bring-up the PortChannel in sonic with LACP and ping between the interfaces. I had to modify the vpp linux-cp plugin to punt/inject lacp packets + modify SAI for IP config and forward packets from tap to PortChannel. Currently in prototype version, I intend to work on refining it in coming days.

[yue-fred-gao]

Hi @AkeelAli, it is great you got it working. I think it might be a good idea to write a HLD about your approach. We can get some input from community, which may help in the final solution.

[AkeelAli]

Sounds good, I will start an HLD to document this prototype version and raise a PR.

[bendrapubalareddy]

@AkeelAli, lets get this committed before LACP change come in. It was tested code but because of lack of time from my side, did not commit. #91 Thanks, Bala

…

On Mon, Dec 16, 2024 at 7:12 AM yue-fred-gao ***@***.***> wrote: Hi @AkeelAli <https://github.com/AkeelAli>, it is great you got it working. I think it might be a good idea to write a HLD about your approach. We can get some input from community, which may help in the final solution. — Reply to this email directly, view it on GitHub <#140 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADKZN6SAV2IFRNSVXDDGDCL2F3UWFAVCNFSM6AAAAABSPNKV6SVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDKNBVHA4DQNBSGI> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[AkeelAli]

@bendrapubalareddy Yes agreed. The LACP changes build on top of #91 so I think it can be committed. I will approve from my side. Note: I had to adapt them to latest as there were conflicts.

[AkeelAli]

Hi @linqingxuan, I added the current LACP design details in #146. Will be adding to this PR as I continue my work. Please let me know how you would like to contribute.

[linqingxuan]

Hello, I apologize that due to work reasons at the end of the year, I haven't paid attention to VPP for a long time. I'd like to know if there are any tasks now that I could help with

[yue-fred-gao]

Hello, I apologize that due to work reasons at the end of the year, I haven't paid attention to VPP for a long time. I'd like to know if there are any tasks now that I could help with

hi @linqingxuan, thanks for offering help. If you are referring to LAG, @AkeelAli can share the current status. If you are talking about sonic-vpp in general, we are focusing on running sonic-mgmt T1 topo test with sonic-vpp. Have you run sonic-mgmt test before?

[AkeelAli]

Re: LAG status: https://github.com/AkeelAli/sonic-platform-vpp/blob/patch-2/docs/HLD/vpp-lag.md#status

[linqingxuan]

Hello, I apologize that due to work reasons at the end of the year, I haven't paid attention to VPP for a long time. I'd like to know if there are any tasks now that I could help with

hi @linqingxuan, thanks for offering help. If you are referring to LAG, @AkeelAli can share the current status. If you are talking about sonic-vpp in general, we are focusing on running sonic-mgmt T1 topo test with sonic-vpp. Have you run sonic-mgmt test before?

Having used sonic-mgmt, I find that I prefer a more hardware-centric approach utilizing Spirent traffic generators, physical vSwitch machines, and GNS3 simulators. This allows me to avoid potential compatibility concerns associated with software-based solutions like sonic-mgmt。If possible, I'd be happy to help with the relatively simple task of code testing. The following tests can be performed:

Sonic VPP on both ends.
Tofino on one end and VPP on the other.
Huawei on one end and VPP on the other.
MLAG on the remote end and PortChannel on the local VPP.
Some other tests.

I made a mistake when submitting. Can the administrator help me reopen it? I'm very sorry.

[linqingxuan]

I will first complete a version of sonic-mgmt tests

[yue-fred-gao]

hardware based testing is definitely valuable. But vpp doesn't have much regressions. On the other hand, sonic-mgmt has lots of test cases and running sonic mgmt did uncover lots of issues. For example, issue #135 was found during sonic mgmt tests. I haven't got time to look at it. If you have some time, maybe you can start with that. You might be able to reproduce it by rebooting manually.

[linqingxuan]

I will try to reproduce the issue and fix it. Has anyone set up our Azure Sonic-img pipeline yet? If not, I'd like to give it a shot.

[yue-fred-gao]

I have a PR for a while but it is not merged yet: sonic-net/sonic-buildimage#21188. I will ping them. BTW, if you update platform/vpp submodule to the latest of sonic-net/sonic-platform-vpp, you need to patch above PR.