Help neededd for cluster dimensioning

[davidefe82]

Hello to everyone.
A customer will change his server soon.
Now he run 2 phisical server, a win 10 and a suse.
He wants to virtualize it in a cluster to have a second chanche in case of failure.
He need a total of 1tb of space.
I think that proxmox ha with ceph is the way. I think that 3 hp ml350 with 2 1tb ssd each is ok to have enaugh space.
I have some question:
It's safe to deploy ha with 3 nodes?
With 3 nodes with 3 total 1tb ssd how much available space will be?
Is there alternatives to have 2 server syncronized realtime?

The customer brutally asked this:
I want to have my vms online on a server, if i unplug ac220 to this server (fault simulation) want to continue working on server 2.
He received an offer by a concurrent based on vmware, he would deploy 2 host with a 3rd server that sync realtime, if server1 fault the vm start on server2, if controller fault the work continue, i don't really understand what is the vmware feature that can do this...

Many thank for your help.

 

[H4R0]

It's safe to deploy ha with 3 nodes?

sure 3 node is the minimum, it will handle 1 node failure

you can get by with 2 nodes and a qdevice (raspberry pi)

With 3 nodes with 3 total 1tb ssd how much available space will be?

well if you use ceph synced to 3 nodes you get 1tb total.

Is there alternatives to have 2 server syncronized realtime?

for realtime sync which is called hyper converged there is ceph and glusterfs

The customer brutally asked this:
I want to have my vms online on a server, if i unplug ac220 to this server (fault simulation) want to continue working on server 2.
He received an offer by a concurrent based on vmware, he would deploy 2 host with a 3rd server that sync realtime, if server1 fault the vm start on server2, if controller fault the work continue, i don't really understand what is the vmware feature that can do this...

they use a shared storage setup, 2 worker nodes, 1 san


to get good performance with hyper converged setups you need ssd storage and >=10G redundant networks.

cost and performance wise I would recommend you to go with shared storage or replication.

also see https://pve.proxmox.com/wiki/Storage

 

[davidefe82]

Hi,
We ask to customer their requirement.
They want to put 3 or more server in 2 different buildings and in case of AC failure in a building wants to remain online.
If i put 2 server in a building and 1 or 2 in other building i dont reach quorum if a building fail.
Is there any way to solve this issue?