New Server Time

[LeMarchand]

Currently running 2012 on a single physical server with a number of VMs that has had most, if not all, of it's disks replaced over the years. More disks getting dodgy, so going to get a new (to us) box.

 

Lots of questions (in bold) to follow!

 

Basically, going to do a similar setup to the existing one: 2 SSDs RAID 1 to run the OS and possibly the VEEAM VM on with the other disks in RAID 5 for everything else. Does this make sense?

 

Next up: 2022 or 2019?

 

Time at that site is VERY limited, so I was considering cheating and just copying the existing VMs as VEEAM replicas in case the old machine dies - that way everything should be running from Day One. Feasible?

 

Finally, I was planning on setting up 2022 (or 2019) "fresh" VMs and moving data and roles across at my leisure. Eventually there will be the same sort of setup but with all the servers on 2022 apart from probably the SIMS server that will remain as a replica (no longer use SIMS) and plan to remove it completely after 5 years. Or should the SIMS data be retained longer?

 

Does the slow replacement idea sound viable?

 

Once everything is hunky dory, the old server would have it's dodgy disks removed and used for offsite backup.

 

Expect a lot more questions in September!

[3s-gtech]

Don't use RAID 5. It's slow. All of my VM hosts use RAID 10 for that reason. You can also let Windows manage it and just pool the disks.

 

I'm currently updating all of my VMs to 2019 from 2012 R2, but really 2022 will be fine and I'd recommend going to that. It's similar enough.

 

You can keep the replicas like that, sure. Hyper-V will just run the VMs that you've moved.

 

You'll need to look at data retention regulations. If everything is in your new MIS, you won't need to retain the old data for as long - but some sensitive data has to be kept longer (I can't remember the specifics). It's 25th birthday of the students in many cases.

 

Your slow replacement will work fine, and is time flexible.

[stevewill6]

Here we use one RAID10 array then partition the array for OS Drive and Data Drive.

 

You don't mention what the other drives are so RAID 10 may not be an option. You want the VMs on the Fastest storage not the OS.

 

When i upgraded my servers i used Hyper v Move to move the VMs to the new server. Worked well.

[KK20]

Ironically with newer server you might end up with BOSS cards that have fast M2 drives, they tend to hold the OS and could well be the fastest drives in the server. RAID 5 on HDDs give me the heebie jeebies, RAID 5 on SSDs isnt too much of an issue and when they are in warranty you dont care about excessive writes. You dont even need to VEEAM replica, if you are creating a new server then have the server migrate the VMs to the new server for you, you could even keep the old server around as a last-ditch cold hyper-v replica (Im not a fan of hyper-v replica, I have had one fail), however if your disaster recovery strategy is "restore from backup to new iron" rather than "high availability with restore from backup as last resort" then you have nothing to lose.

 

Since this server is a hypervisor then 2022. If you ensure you have free space then you can migrate your guest VMs on whatever schedule you like.

 

As for viability, this seems perfectly normal. Effectively you are creating a new hypervisor, migrating VM locations, decommissioning old hypervisor, migrating guest VM OS.