Proxmox geo-distributed cluster, which DFS to take for Encrypted LVM?

Question

LVM

Good day dear community.
There is the following task.
Set up a geographically distributed cluster on the Proxmox platform of 3-4 servers located in different DCs.
There were several solutions on this topic: since a Proxmox cluster can only be made within the same local network, a VPN rises between servers (OpenVPN, mGRE tunnels, other options).
Further, a cluster is created on this and theoretically quite rulitsya.
Here's what I found (perhaps the links will also be useful to someone):
1. Configuring a Proxmox VE 2.x cluster running over ...
2. Proxmox + Ceph virtualization
3. Poor Man's Proxmox Cluster
But the question arose, what file system to take for storage?
Considered Ceph, DRBD, GlusterFS. But my main gag is that on all nodes the disks are driven into a soft raid and encrypted LVM is already installed on it. And as I understand it, for the same Ceph, it is necessary to allocate the entire disk, and picking up LVM Volume is problematic. Or I didn't fully understand the documentation.
As such, the HA cluster does not need to be configured.
That is, the task is to have a complete copy of each virtual machine on each node with some slight delay. And if necessary, it was possible to raise it on another node.
That is why I abandoned the idea of ​​simply replicating backups via rsync, for example.
Does anyone have experience with such an implementation? I would appreciate any advice. Maybe it makes sense to reconsider the very concept of organizing such a cluster?

Answer 1

GlusterFS. It can work on top of anything. I use it in production, while I'm satisfied.
If the cluster is in different DCs, then you will need a good channel, you will have to tune tcp and glsuter itself.
Right, such a Gluster-LVM-raid layer is not good.

Answer 2

You need to extract the variables, but they are also dynamically stored in memory

Not so "dynamically", the process memory structure is perfectly readable.
A lot is done by functions (especially with OOP), you can put hooks on them, esp. Android, etc., where even a table of function names is stored.
What exactly to read = decompilation and debugging.

Answer 3

well, do WH - often the server transmits the coordinates of all the nearest players. Make their display, calculation of their speed, auto-guidance. The server will notice the substitution immediately - and the anti-cheating software will issue a global ban. And if you just listen to the traffic, the chance of being detected is much less. It's better not to touch the process of the game itself - anti-cheating software monitors the introductions into the process and the access of other applications to the process's memory. And if you turn MITM at the kernel level (network adapter driver) - by intercepting traffic without changing it.