HCI (Scale, Starwind etc): vm layout
-
Backup target would go off cluster. Other than that, it would be rare to do an HA cluster but not put all the workloads on it. There is a legitimate concern to have high criticality functions doubly redundant, but practically, that gets pretty silly. Even in the enterprise AD rarely is so important as to be worth working around cluster risks. The cost and complexity of protecting against something so trivial gets pretty high when a restore would only take a few minutes in case of a full cluster failure.
-
How of course, you still run 2+ AD nodes on your cluster, you don't let them failover at the platform level, and you ensure that they stay on different hosts. But you don't put them on a second cluster.
-
If you were to start having AD, databases, etc. on multiple clusters, you have to start asking what the point of the original cluster was
-
@scottalanmiller said in HCI (Scale, Starwind etc): vm layout:
If you were to start having AD, databases, etc. on multiple clusters, you have to start asking what the point of the original cluster was
True dat
-
If you don't have multiple clusters, you obviously weren't sold enough!
-
@FATeknollogee said in HCI (Scale, Starwind etc): vm layout:
For those of you with 2-3 node Scale, Starwind type HCI: are all your VMs run in the cluster or do you run a service like AD or backup vm on a separate host?
We've taken to setting up a dedicated pair of domain controllers in the node's local Hyper-V setup as opposed to making them highly available in our Storage Spaces Direct and disaggregate (Nodes + shared SAS JBOD(s)).
On top of that, with the number of folks getting hit by encryption events we've started to keep the cluster AD separate from the company AD operating the environment the same way we do our cloud tenant infrastructure.
-
@PhlipElder said in HCI (Scale, Starwind etc): vm layout:
We've taken to setting up a dedicated pair of domain controllers in the node's local Hyper-V setup as opposed to making them highly available in our Storage Spaces Direct and disaggregate (Nodes + shared SAS JBOD(s)).
Setting up the DC in the Hyper-V partition?
-
@FATeknollogee said in HCI (Scale, Starwind etc): vm layout:
@PhlipElder said in HCI (Scale, Starwind etc): vm layout:
We've taken to setting up a dedicated pair of domain controllers in the node's local Hyper-V setup as opposed to making them highly available in our Storage Spaces Direct and disaggregate (Nodes + shared SAS JBOD(s)).
Setting up the DC in the Hyper-V partition?
We set up a RAID 1 SSD pair for each host node. The partition created would be carved up to leave 60GB or 75GB free that would then be used for the guest VHDX file hosting the domain controller's OS.
I'm not sure what "Hyper-V partition" means.
-
@PhlipElder said in HCI (Scale, Starwind etc): vm layout:
I'm not sure what "Hyper-V partition" means.
I meant to say the host o/s partition.
-
@FATeknollogee said in HCI (Scale, Starwind etc): vm layout:
@PhlipElder said in HCI (Scale, Starwind etc): vm layout:
I'm not sure what "Hyper-V partition" means.
I meant to say the host o/s partition.
I kinda gathered that is what it meant. Thanks for clarifying.
