Is 802.3ad bonding still not supported for corosync?

[silvered.dragon]

I'm building a new proxmox cluster and I want to use MLAG + separated VLANS for ceph, lan and corosync. Everything it's working, linked and pingable but I'm facing random errors only in my corosync network similar to

[KNET ] host: host: 3 has no active links 802.3ad bond
 [TOTEM ] Retransmit List: 8 9 a 1 5 7 12 13

this happens especially when I reboot nodes, or when I reboot the switches or sometimes random. So I searched a lot and I found this of y 2019

https://forum.proxmox.com/threads/cluster-lacp.90668/

it says that in this https://pve.proxmox.com/pve-docs/pve-admin-guide.html#sysadmin_network_configuration documentation is reported

If you intend to run your cluster network on the bonding interfaces, then
you have to use active-passive mode on the bonding interfaces, other modes
are unsupported.

but I think that the documentation changed because I cannot find this reference. So I want to know if this is still valid, and why LACP is a problem since I think is a lower layer and therefore cannot affect clustering?

 

[sterzy]

Hey,

yeah, this was changed fairly recently [1]. Ideally for Corosync you don't want a bond, you should just hand over the interfaces to Corosync directly. It will handle switching the links better than if you give it a bond.

[1]: https://git.proxmox.com/?p=pve-docs.git;a=commitdiff;h=4ab400d1

 

[spirit]

I'm running proxmox corosync on lacp bond since 12years without any problem

 

[silvered.dragon]

Hey,

yeah, this was changed fairly recently [1]. Ideally for Corosync you don't want a bond, you should just hand over the interfaces to Corosync directly. It will handle switching the links better than if you give it a bond.

[1]: https://git.proxmox.com/?p=pve-docs.git;a=commitdiff;h=4ab400d1

thank you dear, after your usefull information I checked again my mlag environment to understand if was something wrong on switch side, it comes out that with mikrotik if you use more than one bridge, the second one will use cpu for switching so no hardware offloading features, and this was creating latency inside corosync.

I fixed creating a unique bridge with vlan bridge filtering and now is working super good.

Obviously the old documentation was making me thinking on proxmox side.

Thank you