Heartbeat network in a windows 2016 cluster

[sandroalvesbras]

Hello,

I have many servers on a blade where all have dSwitches with VLANID for network production servers, ratification, development and DMZ and so on.

I am creating a windows 2016 cluster for sql server in this VMware environment.

How would I configure a heartbeat NIC in this scenario?

Generally when we have a VMware server and we create a cluster in it for validation, as vSwitch is on the same computer, when I create an isolated vSwitch, it works perfectly.

But as I have multiple servers in VMware clustered with DRS, I was in doubt on how to do this setup.

Or even, do I need a heartbeat NIC for a clustered sql windows?

Thank you.

[ThompsG]

Hi sandroalvesbrasil,

While it is recommended to have a seperate network adapter (read network) for cluster-to-cluster communications and it should be configured so that nothing else talks over it. The cluster communication can also happen over the production network. The reason for a “private” network is so that these communications aren’t impacted by other traffic.

At at this point if your physical network adapters are connected to the same Distributed Switch, then I’m not sure there is much benfit of having a seperate virtual adapter for this traffic. Maybe if you did something fancy with Network IO Control?

Also have you considered Always-On? Assuming here you were talking “Standard WFC” vs something else

Kind regards.

[sandroalvesbras]

Got it.

I have not yet worked with a distributed switch. My experience is only with vSwitch and VLANIDs.

But in this scenario someone configured with a distributed switch.

I stay in doubt if it really was needed, because the current cluster does not have the heartbeat network.

And even if I wanted to configure it, I would have to change the network settings of the environment, creating a VLAN for that purpose at first unnecessarily.

Regarding SQL AlwaysOn, we studied the possibility, but we understood that it would help in the security layer of data integrity and would be a complement to the WFC.

VMware Cluster protects the hardware layer while WFC assists in the protection of the operating system layer and AlwaysOn in data integrity. That's how we thought.

Thank you.

[ThompsG]

Understand however from a network access perspective its not really an issue of Standard vs Distributed though with Distributed you have better control over certain aspects.

In regards to the cluster-to-cluster communication, as long as you provide redundancy for the communications and the heartbeat is able to get through, there are no issues. In our case we are running 2 X 10 Gbps for the VM traffic so going to the extra effort of creating a separate Network (VLAN) to pump this traffic across seemed unnecessary. Your mileage may vary but its worked fine for us for quite a while.

In regards to Always On - your thinking is correct. In a cluster you have a shared volume and if this gets corrupted/deleted/whatever then your data is offline. With Always On you have a second copy of the data online. Doesn't prevent issues within the database but does at a disk level.

Kind regards.