PBS vs 2nd Node

[inxsible]

I have a proxmox server that has 17 containers and 1 VM. I now have in my possession a HP T730 thin client with 32GB SSD and 4 GB RAM. Assuming I upgrade the m.2 SSD to 256 or 512GB and increase the RAM to 8 or 16GB, would it be better to have PBS installed on it and backup the containers or would a 2nd node of live proxmox be more beneficial? In other words, backups of containers/VMs vs a 2nd node of live containers/VMs

I am using this in a home environment, so uptime is not of massive concern even though it might be nice to have during upgrades especially in case something goes wrong with it during an upgrade. Having never used PBS, I am trying to find out how the backups are stored on PBS.

 

[inxsible]

Any tips, comments opinions ?

 

[Dunuin]

I would use it for PBS and then the 4GB RAM should be fine. If downtime is not a problem HDDs for the PBS will be fine even if PBS was designed with only SSDs in mind. But keep in mind that PBS will store everything as small chunks, so a 512GB PBS datstore will store atleast 128000x 4MB files and these all need to be checked at each GC, verify, prune task so HDDs will be very slow because they can't handle IOPS well.
You can also install PBS + PVE to the same machine. So if you upgrade your RAM to 8 or 16GB you could use also restore guests from your PBS to your PVE on the same T730 so your most important guests could still run even if your big PVE server is down.

 

[inxsible]

I would use it for PBS and then the 4GB RAM should be fine. If downtime is not a problem HDDs for the PBS will be fine even if PBS was designed with only SSDs in mind. But keep in mind that PBS will store everything as small chunks, so a 512GB PBS datstore will store atleast 128000x 4MB files and these all need to be checked at each GC, verify, prune task so HDDs will be very slow because they can't handle IOPS well.
You can also install PBS + PVE to the same machine. So if you upgrade your RAM to 8 or 16GB you could use also restore guests from your PBS to your PVE on the same T730 so your most important guests could still run even if your big PVE server is down.

Thanks for your reply. The thin client only takes in SATA m.2 and cannot fit a 2.5" disk. So it will have to be an SSD. However due to the small file size I wonder if it would adversely affect the life of the SSD.

I also have another Dell Optiplex 755 SFF (Core2Duo E8200, 8GB RAM) lying around but the PSU is not working. That could be another option if I buy a PSU for $20 off of ebay. That can fit 1 3.5" hdd or 2x 2.5" hdd/ssds.

What size of drive would you recommend for the PBS -- assuming that I continue using the 17 containers and 1 VM and probably add 1 or 2 additional containers in the future?

Would you please elaborate on running PBS+PVE -- do you mean running PVE and adding a VM or container for PBS? Wouldn't that make it difficult to restore if something were to happen to the PVE itself?

Or did you mean adding a node to PVE and then running PBS on both nodes as VMs----that way we have 2 instances of live VMs and 2 instances of backup ? Maybe that is too much redundancy !! LOL

What are the pros and cons of 2nd pve node vs only PBS ?

A single disk Proxmox node will work for me as I also have a TrueNAS server and I mount the NFS shares on all containers that need it. So the proxmox local storage disks will only be used for OS and container/VM disks.

Thanks again.

 

[Dunuin]

What size of drive would you recommend for the PBS -- assuming that I continue using the 17 containers and 1 VM and probably add 1 or 2 additional containers in the future?

That depends on the size your main PVE node is using and how much backups of each guest you want to store. PBS will use zstd compression and deduplication so its way more space efficient than vzdump backups. With PBS you basically only need to store data once, no matter how often that data is used. Lets say you got a 10GB VM and its data only changes 100MB each day. You first backup of that VM will consume the full 10GB but each following daily backup will only need 100MB (the amount of data that changed) because the other 9,9GB are already there and don't need to be stored again. So if you want to store 14 daily backups of this 10GB VM this would only result in 10GB + 13x 100MB = 11,3GB instead of 14x 10GB = 140GB that a vzdump backup would need. And this is all without compression, so everything actually needs even less space.

Would you please elaborate on running PBS+PVE -- do you mean running PVE and adding a VM or container for PBS? Wouldn't that make it difficult to restore if something were to happen to the PVE itself?

You can run PBS in a LXC, VM or bare metal. I would install PVE + PBS bare metal (install PVE from the PVE ISO, add the PBS repos and then add PBS to it using apt update && apt full-upgrade && apt install proxmox-backup-server) you can then access PVE using port 8006 and PBS using port 8007. Ofcause, it doesn't make sense to run guests on that T730 because you can't back them up on your PBS because if the disk dies you loose your VMs and backups. But my idea was that you don't run guests on your T730 PVE. Its just used for backups but in case another PVE server dies you got an already running PVE where you could temporarily restore your backups to so your most important guests can continue running while you repair/setup your main PVE server.

Or did you mean adding a node to PVE and then running PBS on both nodes as VMs----that way we have 2 instances of live VMs and 2 instances of backup ? Maybe that is too much redundancy !! LOL

That would work too. You could run PVE + PBS on both of them so they could backup each other so you always got a backup and working node in case one of the node dies.

 

[inxsible]

Thank you again.

Most of my containers only run a single service and are only 4GB except the nextcloud and bitwarden which are 16GB & 8GB respectively. As I mentioned, I have a NAS that I use for data, so the OS drives don't have to be large. In total I have 176GB of LXCs+VM. I could put in a 500GB HDD that I have in the Dell Optiplex and use that for PBS until it dies and then replace with a SSD at that time.

the T730 thin client can probably be used as a backup opnsense until I find a better use for it.

 

[Dunuin]

the T730 thin client can probably be used as a backup opnsense until I find a better use for it.

You can also virtualize the OPNsense inside a VM on a PVE . Works fine here in HA mode except for suricata that got problems with KVM.

 

[inxsible]

You can also virtualize the OPNsense inside a VM on a PVE . Works fine here in HA mode except for suricata that got problems with KVM.

Yes, I know. But I like my network to be up in case I screw up a PVE upgrade or something. So opnsense is on bare metal for me.

 

[Dunuin]

Yes, I know. But I like my network to be up in case I screw up a PVE upgrade or something. So opnsense is on bare metal for me.

Thats why I run two virtual OPNsense VMs in HA mode on two nodes. Once one is down the other one will take its role within a second so there is zero downtime even if I shutdown a server. It don't even drop running downloads when the VMs are switching over because OPNsense uses pfsync protocol to be always in sync.
And you even don't need a HA PVE cluster. HA is done by the OPNsense guests themselve:
https://www.thomas-krenn.com/en/wiki/OPNsense_HA_Cluster_configuration

 

[inxsible]

That's interesting. I have never done HA with Opnsense/pfSense before... so it might be nice side-project with the HP T730 thin client or an LXC container