StarWind Software

Hi,

I'm starting with a POC build for a 3-node cluster on Hyper-V with 2 full replication nodes and a third node configured just as a witness (with Node Majority set as failover strategy) and I have a question about the network interfaces on the witness node.

All 3 nodes are connected through 2x Stacked switches.

The 2 main nodes each have 8x10G ports configured into 4xTeams (Starwind Sync, iSCSI, Hyper-V Cluster/Migration, LAN).

The server I'll be using as the witness is however lower spec machine with only 2x10G + 4x1G interfaces.

Seeing as no data actually needs to be replicated to/from the witness node and it a.) won't be an iSCSI target and b.) won't have any VMs migrated to it during failover is there still a requirement for it to have independent links for each service or would it be acceptable to use just the 2x10G links in a team onto the 2 switches for both Starwind sync and Hyper-V management and then 2 of the 1G links for LAN?

Thank you for any advice you could give

Hi,
First of all, we do not recommend any form of NIC teaming for StarWind ISCSI and synchronization interfaces. It is better to use connections on separated NICs via 2x Stacked switches in your case. You can use NIC teaming for Management and VMs traffic.

Seeing as no data actually needs to be replicated to/from the witness node and it a.) won't be an iSCSI target and b.) won't have any VMs migrated to it during failover is there still a requirement for it to have independent links for each service or would it be acceptable to use just the 2x10G links in a team onto the 2 switches for both Starwind sync and Hyper-V management and then 2 of the 1G links for LAN?

You are right, just the same recommendation, without NIC teaming. It can be even 1 GB NICs.