[TUTORIAL] Backup / Restore / Consistency_check

[roan-jr]

Hello all.
Sorry for my English, I'm not a native speaker )
I`d like to share my Proxmox VM backup/restore/consistency_check bash script.
I do it for two reasons:
- it could be useful for someone
- I`d like to get some critique about it. Perhaps this is fundamentally wrong way and you all use another methods.

So, how does it work.
We have a main Proxmox server and a backup Proxmox server, which is ready to run vm's if the main server is down.
(Okay, Proxmox has a replication feature and that's great, but it needs ZFS. And ZFS has noticeably lower performance and much more RAM requirements)

1. We make a VM backup via default Proxmox feature: Datacenter - Backup
2. Copy .lzo to backup server via scp by cron
3. Restore VM on backup server
4. Edit VM config file:
- disable network interface to avoid IP address conflicts when 2 copies of VM are running
- disable cdroms
- disable start on boot
- cut memory (our backup server is not as powerful as a main server)
- insert current date in the VM's name, so we can know when the copy was restored
5. Start VM
6. Check if VM was booted. We could ping it, but the network is already down. So we use qm agent ping.
7. If VM successfully started then we shut id down and remove .lzo file
That's it.

So, this is simplified version of script which is running by cron on a backup server:

### VM Restore function ###
function restoreVMID {
  qm stop $1
  sleep 20s
  qmrestore /var/lib/vz/dump/vzdump-qemu-$1-$(date +%Y_%m_%d)*.vma.lzo $1 --unique 1 --storage local-zfs --force 1
   if [ $? -eq 0 ]
    then
      echo -e "VM" $1 "Successful recovery"
    else
      echo -e "VM" $1 "Recovery failed!"
      return 123
  fi
### Edit VM config ###
  sed -i 's/firewall=.*/firewall=1,link_down=1/' /etc/pve/qemu-server/$1.conf
  sed -i '/cdrom/d' /etc/pve/qemu-server/$1.conf
  sed -i '/onboot/d' /etc/pve/qemu-server/$1.conf
  sed -i 's/memory:.*/memory: 4096/' /etc/pve/qemu-server/$1.conf
  sed -i 's/\(.*\)name:\ /\1name:\ '$(date +%Y.%m.%d)'\-/g' /etc/pve/qemu-server/$1.conf
}

### Let`s count backup quantity, restore our VM`s & test availability ###
countLZO=0
for i in `ls /var/lib/vz/dump/*$(date +%Y_%m_%d)*.lzo | cut -c 30-32 | awk '{print $1}' | uniq`
  do
    restoreVMID $i
      if [ $? -ne 0 ]
        then
          echo $(date +%Y.%m.%d\ %H:%M) "VM" $i "Recovery failed!"
          continue
      fi
    qm start $i
    sleep 3m
    qm agent $i ping
      if [ $? -eq 0 ]
        then
          echo -e "ping" $i "is OK"
          rm /var/lib/vz/dump/vzdump-qemu-$i*.vma.lzo
        else
          echo -e "VM" $i "Ping error!"
          qm stop $i
          continue
      fi
    qm stop $i
    countLZO=$((countLZO+1))
  done
echo -e "Total backup count =" $countLZO$


### Let`s count existing VM quantity ###
countVM=0
for i in `qm list |  awk '{print $1}'  | awk "NR!=1"`
  do
    countVM=$((countVM+1))
  done
echo  -e "LZO" $countLZO "VM" $countVM
echo  -e "FINISH"
exit 0

 

[Dominic]

Hi!

You could mark your thread as tutorial by editing your post and setting a prefix to your title

2. Copy .lzo to backup server via scp by cron (...) 7. If VM successfully started then we shut id down and remove .lzo file

That means the original backup is not on your backup server, but a copy (with edited config) of your original VM is? Consequently, you do not have backups/copies of different points in time but only of a single one?

 

[roan-jr]

You could mark your thread as tutorial

Thank You!

That means the original backup is not on your backup server, but a copy (with edited config) of your original VM is? Consequently, you do not have backups/copies of different points in time but only of a single one?

Quantity of .lzo files on a backup server depends on your storage size and if You have enough space You can leave as many copies as you need.

And what about changed config? Is this a problem to change this settings back after restoring backup on the main server?

 

[Dominic]

? Is this a problem to change this settings back after restoring backup on the main server?

Theoretically or rather if you are sure that there are no errors: I don't think so.