@emad-r said in Linux NiC Bond Question:

Hi,

after you create linux nic bonding, do you disable the original nic adapters, so only the bond network adapter gets IP and has priority.

Cause the only way I saw activity of my nm-bond interface was when I disabled the ens33 and ens34 from getting an IP, so only the bond has it.

Or you keep all interfaces as is with the machine having 3 IPs, how then it will determine to use the bond IP and not others?

I used rr or round robin

I use only use the virtual NICs as well on Debian as well.

@tim_g said in Hyper-V Network card setup?:

@dave247 said in Hyper-V Network card setup?:

Thanks. And this isn't for testing. I actually want to use this server for some production servers.

Sorry, didn't know. I mistakenly assumed testing and lab because of the time frame.

But you can't go wrong with doing it that way.

If you have no need for an extra NIC you could do a 3-NIC team. If you are fine with no management NIC, do a 4-port team and share it with your management OS if for some reason you need that many NICs in a team. Though you probably don't.

Yeah well it is for some low-risk servers, so, kinda almost testing, but not really, if that makes sense. lol

@Mike-Davis said in switch and NIC teaming in Hyper-v:

Good points. Now that I think about it, in this case it's pointless since they have a 1GB uplink to the main switch. Since all ports go in to one switch and there is only one 1GB uplink nothing is gained in terms or redundancy or bandwidth.

You always gain redundancy you can have 3 links fail.

@JaredBusch said in NIC teaming on Hyper-V:

Then you make you vSwitch. If you already have your vSwitch setup, make a team with the ports NOT on the vSwitch, move the vSwitch to the team and then add the final NIC to the team.

Not much to add here. SwitchIndependent mode is a big one on Hyper-V. Sure, Windows can easily use LACP and other means, but what if you want to use two or more uplink switches for redundancy? LACP can't handle this and there is just a handful of proprietary protocols that can. SwitchIndependent mode is doing exactly this by "load balancing" VMs and Host traffic between the available links and failover in case something goes south.

This way, like @JaredBusch said above, you can have LACP-like functionality (max single port speed for a single traffic source) over multiple inexpensive switches. In fact, the switch doesn't know anything about that type of teaming, you could even use unmanaged switches (but really, don't do that)

My hosts are running in this mode.