OS NIC lost link detection when bridge interface is configured

[vercetty92]

Hi,

I have a strange thing happening on a blade bl460g9 inside an enclosure hp c7000.

it only happens to one blade (the first blade. OS is CentOS7).

I have 2 bonding interfaces active-backup mode. bond0 have a bridge NIC and works fine.

bond1 have another bridge NIC, but impossible to have this one working fine.

Everytime I mount bond1 + bridge-bond1 : I lost the connection on the 2 NIC that compose bond1 (eno53/eno54), I can see that with ethtool.

If I remove the bridge NIC and mount only bond1 with an IP + reboot, no pb at all. bond1 is working fine.

I don't understand because bond0 don't have this pb.

sometime even a reboot + single bond1 don't solve the issue, I need to stop the blade, unassign the server profile, assign it again, start the blade, and here it's ok.

But for the moment, impossible to have the bridge working on bond1, as the link are always lost.

I can see that it happens when OS start, because if I do a continuous ping to the bond1-bridge interface, I can ping it a few second after OS is started, then it stop.

Exact same setup work fine on blade 2 and blade 3.

Does someone know something about such behavior?

Thx a lot