Changing NICs from 1Gb to 10Gb in an availability group

Question

We have a two-node Always On Availability Group set up. Both nodes are on SQL Server 2016 SP3.

Primary server has two 1Gb NICs. Secondary has two 10Gb NICs.

Both primary and secondary have a default gateway configured for all four IPs.

We want to enable a single 10Gb NIC on the primary server and disable the default gateway for one of the IPs on both nodes, following these steps:

1. Disable default gateway for one of the IPs on the secondary server
2. Failover to secondary
3. Disable default gateway for one of the IPs on the primary server
4. Configure 10Gb NIC on the primary
5. Disable one of the 1Gb NICs on the primary
6. Failback to primary

Will the above steps be enough to accomplish the task?

Does the cluster automatically identify the new NIC and reflect it in cluster network details?

Can we run the primary with 1Gb NIC and 10Gb NIC (disabling default gateway for 1Gb NIC IP)?

Reasons for doing this:

1. We are upgrading our network from 1Gb to 10Gb, so we want to utilize the 10Gb pipe.
2. We are copying backups from prod to DR server and from there to tapes. A couple of our DBs are around 6TB with backups around 500GB-800GB. We want to speed up the copy too. NICs on these boxes are not used.

Answer 1

1. We are upgrading our network from 1Gb to 10Gb - so we want utilize 10Gb pipe.

Great! Definitely a step in the right direction. There isn't anything extra needed except to make sure you're on the latest firmware and driver set. Also make sure any advanced settings are set correctly, such as RSS. RDMA, configuration, testing, and hardware compatibility is beyond the scope of this site.

2. Also we are copying backups from prod to dr server and from there to tapes. And couple of our DB's are around 6tb with backups around 500g-800g....we want to speed up the copy too.

Going from ~125Mbytes/sec to 1.2 Gbytes/sec should help, assuming there are no other bottlenecks in the system (which would be something a network engineer should look at from the network side).

Now that we know what you're wanting to do, which is replace the 1 Gb nic with a 10 Gbit nic, then that's all you need to do.

Let's get back to your original post.

We want to enable a single 10Gb NIC on primary server and disable default gateway for one of the IP's on both nodes.

The first part is covered, drop the new nic in, take the old one out. For this, your steps are correct in failing over so that you can do maintenance. Depending on the anticipated downtime, you may want to remove that replica from the AG. Whether you do that or not depends on how much log holdup you can endure as an unhealthy database replica will hold the log truncation.

The second part, about the gateway, makes no sense. Gateways are set on the interface, not on an IP, and if you don't have a gateway then the data won't have a default route, which would be silly in this case since you want this nic to talk to other servers most likely outside their own subnet.

Does the cluster automatically identifies the new NIC and reflect it in cluster network details?

Yes, when the new adaptor is installed, clustering will see it and the settings for it and update anything it needs to which would mostly be the interface manager. This is automatic and shouldn't require any intervention.