The best disk configuration for two arrays

[driftux]

Hi, I am planning very serious sever with the HW raid card. I plan to create RAID1 from 2x250GB SSD and RAID10 second array with 6x2TB drives.
Two VM will be created on RAID1. I plan to hold a lot of emails maybe 2000 accounts on one of the VM, so I want export all the /home directory from the machine and mount it on the RAID10 array. Near this server I will put the exact configuration another server which also will have RAID1 from 2x250GB SSD and RAID10 second array with 6x2TB drives and want to do replication from the main server using Proxmox backup server. So having a big failure on the main server I will recover both VM on second server.

So I want you guys to ask:
a) Is the best way to export /home directory and to create a datastore on RAID10, create virtual machine disk in there, format it and attach it to VM?
b) I also read that possible to do disk pass-through configuration. But is it possible in my scenario?
c) Also read in other forum that possible passing the entire SATA controller or RAID controllers to the virtual machine (which would then inherit the disks connected to the controller), but I have no idea, can I do it with Proxmox?

And the main point is that I will use Proxmox Backup server for the replication/backup between those two server, so Proxmox Backup server must understand the chosen configuration. I will kindly wait for your comments. Thanks.

 

[driftux]

Is my questions are to complicated?

 

[ertanerbek]

I think you will use stanalone mode on that two server wthouth cluster, because if you will build serius system with proxmox and full redundancy you must build 3 server.

So if you will be use cluster then do not use raid card, basicly build CEPH ( need 3 cluster for correct work and radundancy )

If you not use cluster and will be build cluster on aplication side then you can use LVM Thin volume..



a) Is the best way to export /home directory and to create a datastore on RAID10, create virtual machine disk in there, format it and attach it to VM?

Where stor that mail server customer mail ? in home I think no it should be stay on /var so 6x2TB I hope that SAS if you serius you will be get 300 IO, is that enought for you ?

b) I also read that possible to do disk pass-through configuration. But is it possible in my scenario?

If you will use stanalone server and do not interasting with migration why not ?

c) Also read in other forum that possible passing the entire SATA controller or RAID controllers to the virtual machine (which would then inherit the disks connected to the controller), but I have no idea, can I do it with Proxmox?

Yes you can but your server CPU must be support VTD also please search pci-passthrough on google




"And the main point is that I will use Proxmox Backup server for the replication/backup between those two server, so Proxmox Backup server must understand the chosen configuration. I will kindly wait for your comments. "

I am not sure but if you will use pci-passthruh I think you can not be get backup with Promox Backup server...

 

[driftux]

Thanks for the answer. I don't want build a cluster, because I want quite easy management.
My HDDs are 5400RPM drives, but I hope that RAID10 will save me. We are talking about email server with 2000-3000 mailboxes.
I understood about pci-through, but my main purpose is that Proxmox backup server would work. So as I understand I can't use passthrough, correct?

 

[bobmc]

What you are trying to do, I think is to have a DR solution - A main server with a continously replicated backup server ready to use if the main server fails. Possible but complicated to setup and manage.

A cluster will make the management as easy as it can be - but the financial cost will be higher. Most commercial setups would use a cluster with possibly an off-site data replication in case of major disaster.

this video gives an overview of a proxmox cluster and handling failure of a node - it's in a lab and not real hardware but it gives you an idea of the process

cluster demo

You don't want to pass the disks directly through to a VM except in very specialised cases, it's a world of potential problems for very little gain

 

[driftux]

Ok, thanks I will look one more time to a cluster solution. By the way what about system performance, let's say I will have Ceph cluster and a plain EXT4 file system server. Which perform better as web server or mail server? It's ok to me to have 1-2 downtimes per year for about an 1-3 hours.
The main factor for me is performance.

Could you please explain in details why do you think I will be in trouble trying to manage my primitive setup of two servers? I think it's very effective solution, easy to control, easy to spread load, easy to solve problems and I can have 20-30 servers and grow more and more. What do I only need is to have more disks space on hdds, for example 30% will be for operable data, and 70% for full backup from the brother server and for example a week long snapshots for every hour. Isn't it so?