Complete Proxmox Backup Server Setup Guide

Before starting, you should have completed the PBS first boot setup (post-install configuration, repository setup, subscription nag removal, and system update).

Step 1: System Optimizations

Before starting, apply some system optimizations that are generally accepted as a good practice for PBS.

Reduce swap usage for better performance (PBS should use RAM):

echo "vm.swappiness=10" >> /etc/sysctl.conf

Disable IPv6 if your network doesn't use it:

echo "net.ipv6.conf.all.disable_ipv6 = 1" >> /etc/sysctl.conf

echo "net.ipv6.conf.default.disable_ipv6 = 1" >> /etc/sysctl.conf

Apply settings immediately:

sysctl -p

Step 2: Storage Setup

Understanding Your Storage Options

PBS requires storage for two purposes:

Operating System and PBS software (~32GB minimum)
Backup datastore (where your actual backups are stored)

Option 1: Single Disk System

Using one SSD for both OS and backups.

Minimum Recommended Size: 256GB SSD (absolute minimum 128GB)

Pros:

Simpler setup
Less hardware required
Lower cost
Good for small deployments (5-10 VMs)

Cons:

OS disk failure means losing backups too
OS operations compete with backup I/O
Can't easily expand storage
Harder to migrate to new hardware

Option 2: Dual Disk System

Separate disks for OS and backups (recommended).

Minimum Recommended Sizes:

OS Disk: 64-128GB SSD
Backup Disk: 500GB+ (depends on backup needs)

Pros:

OS can be reinstalled without affecting backups
Better performance (isolated I/O)
Easy to add/upgrade backup disk
Can move backup disk to new server
Follows backup best practices

Cons:

Requires additional hardware
Slightly more complex setup

Single Disk Setups

For a single disk setup simply create a directory and add it as a datastore.

mkdir -p /mnt/datastore

Then add it as a datastore

proxmox-backup-manager datastore create main /mnt/datastore

Alternatively you could create a dedicated partition, however that adds some complexity without major advantages.

Dedicated Backup Disk

Identify Your Disks

Whether you are using a single disk or two, it is important to understand how your disks are setup. The best way is to list out the block devices on your host

lsblk -o NAME,SIZE,TYPE,FSTYPE,MOUNTPOINT,MODEL

In my case, I have a dedicated OS disk + a dedicated backup disk, and my output looks like:

NAME           SIZE TYPE FSTYPE      MOUNTPOINT MODEL
nvme1n1      476.9G disk                        TWSC TSC3AN512E6-F2T60S
├─nvme1n1p1   1007K part                        
├─nvme1n1p2      1G part vfat        /boot/efi  
└─nvme1n1p3  475.9G part LVM2_member            
  ├─pbs-swap     8G lvm  swap        [SWAP]     
  └─pbs-root 451.9G lvm  ext4        /          
nvme0n1        3.7T disk                        TEAM TM8FP4004T

From this I can see that my operating system is located on nvme1n1. There are a few indicators, first the /boot/efi mount point is on partition 2 (p2) of the nvme1n1. I can also see the lvm filesystem on partition 3 (p3).

The disk I will be using is the 3.7T nvme0n1.

File System

This guide will be using ext4

Mature, stable, well-tested
Good performance for PBS workloads
Simple recovery if issues occur
No additional RAM requirements

The first step will be to wipe the system

I have intentionally omitted the device to prevent accidentally copy and pasting my device. Make sure to fill in your device

wipefs -a /dev/<YOUR DEVICE>

Create a single partition using the entire disk

fdisk /dev/<YOUR DEVICE>

Then in fdisk:

Press 'g' - create new GPT partition table
Press 'n' - new partition
Press Enter to create partition 1
Press Enter for the default first sector
Press Enter for the default last sector (uses whole disk)
Press 'w' - write and exit

You can verify that the new partition was created with

lsblk | grep <YOUR DEVICE>

Format the partition (make sure to include the partition, like nvme0n1p1). I add the -m 0 to set reserved space to 0 as this is not needed

mkfs.ext4 -m 0 -T largefile -E lazy_itable_init=0,lazy_journal_init=0 /dev/<YOUR DEVICE><YOUR PARTITION>

-m 0 = No reserved space
-T largefile = Optimize for fewer, larger files (PBS chunks)
-E lazy_itable_init=0 = Initialize now (better performance later)

Create the mount point

mkdir -p /mnt/datastore

mount /dev/<YOUR DEVICE><YOUR PARTITION> /mnt/datastore

Get the UUID of the disk

blkid -o value -s UUID /dev/<YOUR DEVICE><PARTITION>

Add permanent mount entry to fstab (replace UUID_HERE with output from above):

echo "UUID=UUID_HERE /mnt/datastore ext4 defaults,noatime 0 2" >> /etc/fstab

Test the mount configuration. First unmount the drive

umount /mnt/datastore

Then mount it via the fstab entry

mount -a

Verify

df -h /mnt/datastore

After mounting your storage, create the PBS datastore:

proxmox-backup-manager datastore create main /mnt/datastore

Step 3: Configure PBS

It is always a good idea to try and optimize when things run, here is the schedule I am following:

Optimized Job Schedule:

21:30 - Prune (frees space before backup)
22:00 - Backup (all nodes at once - your fast network can handle it)
01:00 - Verification (after backups complete)
02:00 - GC on Sundays only (weekly is sufficient)

Homelab-Optimized Retention:

3 most recent (quick recovery)
7 daily (one week)
4 weekly (one month)
2 monthly (two months)
0 yearly (unnecessary for homelab)

Configure Pruning

On PBS server - Set retention policy and schedule for automatic pruning:

proxmox-backup-manager prune-job create daily-prune \
    --store main \
    --schedule "21:30" \
    --keep-last 3 \
    --keep-daily 7 \
    --keep-weekly 4 \
    --keep-monthly 2

Homelab-optimized retention:

--keep-last 3: Keep 3 most recent (for quick "oops" recovery)
--keep-daily 7: Keep one per day for a week (most common recovery window)
--keep-weekly 4: Keep weekly for a month (good balance)
--keep-monthly 2: Two monthly backups (sufficient for homelab)
--keep-yearly: This is omitted as we will do not yearly backups
--prune-schedule "21:30": Run 30 minutes before backups to free space

Configure Garbage Collection

On PBS server - Schedule garbage collection to reclaim space weekly:

proxmox-backup-manager datastore update main \
    --gc-schedule "sun 02:00"

Configure Verification

proxmox-backup-manager verify-job create verify-main \
    --store main \
    --schedule "01:00" \
    --ignore-verified true \
    --outdated-after 30

verify-main: Job name
--store main: Datastore to verify
--schedule: Run daily at 01:00 (after backups complete)
--ignore-verified: Skip recently verified backups
--outdated-after 30: Re-verify backups older than 30 days

Create Backup User

On PBS server - Create a dedicated user for backup operations:

proxmox-backup-manager user create backup@pbs --password <YOUR PASSWORD>

On PBS server - Generate an API token that will be used later (copy this):

proxmox-backup-manager user generate-token backup@pbs api-token

On PBS server - Add the DatastoreBackup role to the user

proxmox-backup-manager acl update /datastore/main DatastoreBackup \
    --auth-id 'backup@pbs'

On PBS server - Grant backup permissions to the API token (write-only for security):

proxmox-backup-manager acl update /datastore/main DatastoreBackup \
    --auth-id 'backup@pbs!api-token'

Note: DatastoreBackup role allows creating backups but not deleting them - this protects against compromised nodes deleting existing backups.

Note: You MUST add the DatastoreBackup role to both the User and API token.

Step 4: Connect Proxmox VE

I am running a three node cluster and setting my backups at the datacenter level. Another option for this type of setup is creating a namespace for each node, and setting backups individually per-node. This is useful if you want to apply varying backup schedules depending on the VMs a node has.

On PBS server - Display the PBS certificate fingerprint needed for PVE (copy this):

proxmox-backup-manager cert info | grep Fingerprint

On ANY single PVE node (e.g., PVE0) - This command adds PBS storage for the entire cluster:

pvesm add pbs pbs-backup \
    --server 192.168.1.100 \
    --username 'backup@pbs!api-token' \
    --password 'SECRET-TOKEN-STRING' \
    --datastore main \
    --fingerprint 'XX:XX:XX:XX:...'

You only run this ONCE on ONE node - it automatically becomes available on all cluster nodes!

Parameters explained:

pbs-backup: Storage name in PVE (you can choose any name)
--server: PBS server IP address
--username: PBS user created earlier
--datastore: PBS datastore name (we created "main")
--fingerprint: Certificate fingerprint from PBS

Datacenter-Level Backup (Recommended)

On ANY single PVE node - Create one backup job for all VMs on all nodes:

pvesh create /cluster/backup \
    -id datacenter-backup \
    -schedule "22:00" \
    -storage pbs-backup \
    -mode snapshot \
    -all 1 \
    -enabled 1

This single job will:

Run every night at 22:00 (30 minutes after pruning at 21:30)
Backup ALL VMs and containers on ALL nodes in the cluster
Store everything in PBS's main datastore
Show up in every node's web UI under Datacenter > Backup