What're people using for their Hyper-V hosts?

[fredbayker]

Have you looked into hyper-v replica?

[Pyroman]

Sorry to jump on an old post but I'm just getting my head around virtualisation, we've been running single servers for years.

 

If I get a rack server with enough storage to virtualise our servers do you not use the raid card as a raid and instead have it as a bunch of drives and use storage spaces / Hyper-V as a software RAID?

[Jcx500]

Depends how you want to set it up ?

 

If your just going to have hyper-v on one box it makes sense just to have traditional raid eg

 

2x128gb drives in raid one for os

 

Then another array for hyper-v data

 

Storage spaces is normally when you have multiple hosts in a cluster.

 

I believe you can use it on a single server if you want to but I don’t know many people who do this for single hosts

[Garacesh]

Sorry to jump on an old post but I'm just getting my head around virtualisation, we've been running single servers for years.

 

If I get a rack server with enough storage to virtualise our servers do you not use the raid card as a raid and instead have it as a bunch of drives and use storage spaces / Hyper-V as a software RAID?

 

No worries, you're basically me a year ago Most of the stuff in this thread is probably still applicable.

 

I ended up going with 2 big ol servers, dual Intel Xeons and 192GB DDR4, set up in a cluster with their NVME drives set up in S2D (needs Datacentre, though if you're going Hyper-V you might as well buy Datacentre anyway for the infinite VM allocation).

They're working a treat here. Failover takes around 30 seconds from 'server died' to 'everything's back up and running'

Edit: Just be aware, if you go down the same route, don't make my mistake: you cannot mix SATA/SAS/NVMe drives in S2D, they all gotta be the same type.

 

Not including the Datacentre licenses themselves, was about 7 and a half grand.

 

Edit2: Looking back on my OP, I did end up siphoning the file server off onto a physical box, instead of hosting it virtually. But so far I'm running a DC, an application server, a print server, SIMS, MECM, an RDS server and our backup server on them, and they're chugging along just fine

Edited May 12, 2021 by Garacesh

[Pyroman]

No worries, you're basically me a year ago Most of the stuff in this thread is probably still applicable.

 

I ended up going with 2 big ol servers, dual Intel Xeons and 192GB DDR4, set up in a cluster with their NVME drives set up in S2D (needs Datacentre, though if you're going Hyper-V you might as well buy Datacentre anyway for the infinite VM allocation).

They're working a treat here. Failover takes around 30 seconds from 'server died' to 'everything's back up and running'

Edit: Just be aware, if you go down the same route, don't make my mistake: you cannot mix SATA/SAS/NVMe drives in S2D, they all gotta be the same type.

 

Not including the Datacentre licenses themselves, was about 7 and a half grand.

 

Edit2: Looking back on my OP, I did end up siphoning the file server off onto a physical box, instead of hosting it virtually. But so far I'm running a DC, an application server, a print server, SIMS, MECM, an RDS server and our backup server on them, and they're chugging along just fine

 

Awesome, thanks very much.

 

When you've got the storage spaces set up, does it just appear as one instance of Server 2019 and if you spin up a VM it replicates it onto both servers? Looking at previous posts S2D is basically RAID but with machines rather than disks?

[Garacesh]

It's kind of both, in a way.

 

You group disks together as one logical disk (except it doesn't show up in Disk Management, or get a drive letter, because why would it? That would make too much sense! ) and then mirror those disks, essentially.

So you end up with a folder, in my case, C:\ClusterStorage\$S2DDiskName\ and that gets replicated on both servers. So all my virtual machines, their disks, snapshots, etc are all told to live in subdirectories of that path.

 

Edit: Also worth adding, if you go with 2 servers, you'll need to set up a quorum witness too, just to make sure both hosts don't try to spin up all the VM's thinking the other is dead (there's a name for this but I can't remember what it is..). But that's really easy to do. You can stick that function on any of your existing (physical) servers.

Edited May 12, 2021 by Garacesh

[mavhc]

Always do software raid, unless you raid card costs 1000s it's going to be bad

[Pyroman]

It's kind of both, in a way.

 

You group disks together as one logical disk (except it doesn't show up in Disk Management, or get a drive letter, because why would it? That would make too much sense! ) and then mirror those disks, essentially.

So you end up with a folder, in my case, C:\ClusterStorage\$S2DDiskName\ and that gets replicated on both servers. So all my virtual machines, their disks, snapshots, etc are all told to live in subdirectories of that path.

 

Edit: Also worth adding, if you go with 2 servers, you'll need to set up a quorum witness too, just to make sure both hosts don't try to spin up all the VM's thinking the other is dead (there's a name for this but I can't remember what it is..). But that's really easy to do. You can stick that function on any of your existing (physical) servers.

 

Brilliant, thanks very much :-)

 

Not that there's any spare cash at the moment but re-writing the disaster recovery plan has slightly terrified me