Does GlusterFS + OpenVZ / KVM + Proxmox work in production, are there any problems, do I choose storage systems?

Question

OpenVZ

The question is short:
Does GlusterFS get stuck if machines running KVM and OpenVZ are spinning on it?
Which FS to choose for placing GlusterFS ext3,ext4,xfs,zfs?
Proxmox, a good choice or worth looking at other options?
Full story:
There was a task to deploy a three-node cluster for virtual machines.
I choose a file system for this task.
There is no dedicated server for storage.
You need to store data on the same three nodes.
Ideas so far are two NFS cluster and GlusterFS.
Does anyone have experience with similar solutions?
I read on English-language resources that OpenVZ slows down on GlusterFS (large performance losses), is this true?
And what is the best file system to choose under all this ext, xfs, zfs?
If any other solutions are better, I'll be glad to hear.
I just didn't study clustering issues ovirt , lxcenter.
UPD:
I settled on proxmox because of KVM + OpenVZ support, cluster and a convenient backup mechanism (which I can give into the hands of other users / administrators).
In ovirt, judging by the reviews, the part with backup is lame, lxcenter is based more on xen + kvm.
UPD2:
The main problem is storage. Many solutions are based on a separate storage system or a pair of storage systems, I simply don’t have such. There are 3 physical servers and no more server. They will be loaded by 40-50%, so I can’t give one server for storage :(
UPD3: I’ll
add more information on the disk subsystem.
I will post a description of x-k servers, because this is not relevant to the question.
3 servers. 2 discs each.
I have to make from 2 disks something like RAID1. It doesn't matter what means.
Otherwise, when one disk fails, we lose the entire node.
It turns out that GlusterFS (or something else) comes next. With the configuration of copying everything to all disks.
Why so: If VM1 sits on node1, and its data is on node2, then we get a read / write through the network.
That with an already not powerful storage system, it will clearly put the task to a standstill.
If there is a replica for all nodes, then I get a local VM running on storage + a copy of this VM on other nodes.
Yes, we are losing free space on the HDD, but these are ordinary disks and everyone understands that there will be losses.
500 GB is enough for all three nodes.
Yes, and local recording will allow me to use only two network interfaces (there are 2 of them by default in servers).
1. Shared network 2. GlusterFS synchronization network.
Those. 3. a network for writing to a neighboring node, or better to combine interfaces for GluserFS. not required.

Answer 1

well, it's just that if you have big data chasing to and from disk, then everything will be slow on a 1 gigabit network if you have two proxmox nodes and one nfs node.
If glasterfs, then everything will be there quickly, but it’s not clear why there are three nodes, it’s logical to have two, on one everything works, on the second everything syncs, the first fell on the second everything rose.

Answer 2

Hello. Tell me how are you?
A friend of mine put glusterfs on top of zfs.
This does not give any advantages (even rather the opposite, because zfs is inferior to others in speed tests), but I just like zfs due to the convenience of adding disks and volumes and fault tolerance.
It is not clear whether the deduplication and archiving capabilities of zfs work on the volume of the gluster or not. In theory yes, but in practice

Answer 3

If 3 nodes, then why not RBD?