Hyper-V Cluster with Local Storage questions

[mdrabble]

Our current Hyper-V Cluster has 3 hosts which are approaching 8 years old and our Storage server is nearly 6 years old - so I am looking at replacement options.

 

Host and Storage server all have 10GB Intel SFP+ Nics which will be moved over to the replacement servers.

 

I am thinking the cheaper option is to replace the hosts and storage server with 3 replacement hosts with lots of local storage and configure the Hyper-V Cluster so the hosts use local storage.

 

My questions are......

 

A) If I go down the route, how does the replication side of things work in terms of failover?

B) How does Veeam cope with this kind of setup?

C) Any reason why I should not consider this setup?

 

Cheers

[Jaan]

We use a 3 node S2D solution here.

 

The nodes all replicate to each other, if the case of a node failure it failsover to to another node. Each of the nodes have local storage. SSD cache layer with a spinning rust storage layer.

 

We also use veeam. Whilst its doesn't directly support S2D 2019 (yet), we replicate the Vms from the cluster to a standalone "back up server" which then back ups to it's own independent SAN and replicates that to an "offsite" nas in another building. We then have to tape.

 

We went from having 3 dedicated hyper-v hosts (non clustered), to this setup. The difference is amazing.

 

Would recommend looking at VeryPCs options and Stones options.

[dhicks]

I am thinking the cheaper option is to replace the hosts and storage server with 3 replacement hosts with lots of local storage and configure the Hyper-V Cluster so the hosts use local storage.

 

We've just been through a similar process - moving to new hardware running Server 2019 with local storage. Our original plan was to set up a cluster for instant failover if needed, but it wound up a fair bit cheaper for us to buy just the one new server and re-use a second one as a backup / failover server. We have replicas copied across from the live server to the backup server nightly via Veeam, so if the main server fails we can start up the replicas on the backup servers. Those servers will be a day out of date, but we figure that doesn't matter too much - we could increase the replication period to several times a day.

 

We have Veeam handling both backup and replication of VMs, seems to be working fine so far. Local storage on the main server is a Server 2019 Storage Spaces array including two fast NVMe SSDs. The backup server is linked to a storage unit with a bunch of 12TB drives in RAID6, giving us 70TB of backup space. VeryPC specified the main server for us, complete with warranty and hardware certified to work with Windows Storage Spaces.

[mdrabble]

Current have 27 VMs running across the 3 hosts - 6TB in storage, so I am thinking I still need 3 hosts for the moment as I hoping to combine some services and slim down the number VMs overtime.

 

Did spot some nice HP 1U Gen9 servers on ICT Direct which look like they would do the job

 

 

Edit: --- Just as I am posted the previous bit, the Principal just popped in to ask for a meeting in January to look at the 5 year plan (well they asked for the next 12-24 months) and to talk about budgets etc. - Looks like I am going to be told to tighten the IT Support belt and see where I can cut costs!

Edited December 6, 2019 by mdrabble

[Chris_Cook]

We did this last year and it was great. We got good feedback from our users about how much quicker things worked - especially sims. I don't think we told anyone we were changing servers, but they knew something had been done.

 

We went from a 3 node cluster with SAN and HP servers to 2 dell servers (non-clustered) and local storage + SSD cache. Its a much simple setup which is good.

 

We also were in the middle of our google drive migration - so that drastically reduced the amount of storage we will need in the future.

 

The 2 things we 'lost' were:

 

* no quick migration between hosts - you have to move the VMs storage as well. With a 10g link, this isn't really an issue in practice. For maintenance, it just takes minutes to transfer. For DR, we replicate to our old servers + have backups. Live migration still works well, even during the day.

 

* if a server gets moved to the other host, then you may have to update your veeam job. I think you can just tell it to backup everything from a server, so again probably not a big issue.

 

I would recommend working closely with your server supplier if you want to do storage spaces direct as it has (or had) some specific hardware requirements. It can be expensive if you get the wrong disk controllers etc.

[mbedford]

We use a 3 node S2D solution here.

 

The nodes all replicate to each other, if the case of a node failure it failsover to to another node. Each of the nodes have local storage. SSD cache layer with a spinning rust storage layer.

 

We also use veeam. Whilst its doesn't directly support S2D 2019 (yet), we replicate the Vms from the cluster to a standalone "back up server" which then back ups to it's own independent SAN and replicates that to an "offsite" nas in another building. We then have to tape.

 

We went from having 3 dedicated hyper-v hosts (non clustered), to this setup. The difference is amazing.

 

Would recommend looking at VeryPCs options and Stones options.

 

 

Similar setup here, 6 node S2D 2019 setup, all NVME storage (around 90TB). Works a treat. Also using Veeam for backup and a bit of replication.