Good backup system for LXC / KVM Servers

[sahostking]

Hi all,

I am testing and install a POC box with Proxmox 4 on it.

I'd like to know what is recommended way or what are most of you using for backup solutions of all VMs?

I see it has GlusterFS? Is this the preferred method?

I have 3 test servers all setup in RAID 10 with ZFS partition, ready to roll. I can wipe them at any time Any recommendations on software to go for would be great.

PS: Considering moving to Proxmox from SolusVM. Their backup features are not so good.

 

[ScOut3R]

Well, I use a backup system from the outside. I found the integrated backup solution too slow for us (we are running KVM based VMs only on Ceph storage). So what I do is I have a dedicated backup server, ZFS on it, and using http://www.dirvish.org to do the hard work. I was using dirvish before Proxmox was introduced into the infrastructure. Dirvish is nice, because I can backup almost everything with it, bare metal, remote VPS, VMs on Proxmox.

 

[RobFantini]

we use a stand alone pve server and run nfs on it. normally located on other side of the building.

that way if the entire server room explodes we can restore and run essential vm's on the nfs/pve/backup system

in addition on each pve host and vm we use obnam to make backups to the nfs server and and off site location.
obnam is great for yearly/monthly/daily/hourly/ every 15 minute etc backups. so if vm is restored you can also have recent data backup for the vm.

In addition we use a lot of turnkey systems . they have a fantastic backup and restore system built in. so if the building with nfs and server room are gone then within a short time the systems can be restored to run on amazon ec2 [ i think it is called ec2 ].

 

[yena]

So you install obnam on the Proxmox Host, and please, can you explain me the Restore procedure of a Container Backup ?

Thanks!

in addition on each pve host and vm we use obnam to make backups to the nfs server and and off site location.
obnam is great for yearly/monthly/daily/hourly/ every 15 minute etc backups. so if vm is restored you can also have recent data backup for the vm.

In addition we use a lot of turnkey systems . they have a fantastic backup and restore system built in. so if the building with nfs and server room are gone then within a short time the systems can be restored to run on amazon ec2 [ i think it is called ec2 ].

 

[RobFantini]

So you install obnam on the Proxmox Host, and please, can you explain me the Restore procedure of a Container Backup ?

Thanks!

check http://pve.proxmox.com/wiki/Backup_and_Restore

regarding obnam - we've been using for a few months , it is I think still beta software.
for last 8 years or so have been using rsnapshot .
I think rsnapshot would be easier to start with - and use obnam as a secondary system.

so we use at least 2 data backup systems.

I can post info on how to use later if you are interested.

 

[yena]

I'm using Rsnapshot too but i use it on the host server or in a dedicated container...
The problem with Rsnapshot is MySQL Backup ...
Have you a solution for this problem ?

Thanks!

check http://pve.proxmox.com/wiki/Backup_and_Restore

regarding obnam - we've been using for a few months , it is I think still beta software.
for last 8 years or so have been using rsnapshot .
I think rsnapshot would be easier to start with - and use obnam as a secondary system.

so we use at least 2 data backup systems.

I can post info on how to use later if you are interested.

 

[RobFantini]

I'm using Rsnapshot too but i use it on the host server or in a dedicated container...
The problem with Rsnapshot is MySQL Backup ...
Have you a solution for this problem ?

Thanks!

install automysqlbackup

then include /var/lib/automysqlbackup/ as a directory to back up.

 

[yena]

Like me but i use a custom script .. but a better solution should be a snapshot before Rsnapshot backup..
because if i have to restore many VM like 100 and i have to re-dump all mysql databases for every VM .. should be a too long process ...

 

[RobFantini]

Like me but i use a custom script .. but a better solution should be a snapshot before Rsnapshot backup..
because if i have to restore many VM like 100 and i have to re-dump all mysql databases for every VM .. should be a too long process ...

that is a good point. for us that is not necessary as our mysql data does not change often.

have you considered using zfs snapshots?

 

[yena]

I'm Testing This:
http://pve.proxmox.com/wiki/PVE-zsync

 

[term]

check http://pve.proxmox.com/wiki/Backup_and_Restore

regarding obnam - we've been using for a few months , it is I think still beta software.
for last 8 years or so have been using rsnapshot .
I think rsnapshot would be easier to start with - and use obnam as a secondary system.

so we use at least 2 data backup systems.

I can post info on how to use later if you are interested.

I'd be interested in hearing the details. Right now we just rsync proxmox's backup images, but that is probably not the most efficient way...

 

[RobFantini]

I'd be interested in hearing the details. Right now we just rsync proxmox's backup images, but that is probably not the most efficient way...

I think rsync is a great tool for backing up the vzdumps to another system. I do that over night to a few pve systems containing extra storage. then to get off site we transfer images to usb to bring off site . we use the package usbmount to control mounting of usb. usb 3.0 is a lot faster so if a system only has usb 2.0 we add a usb 3.0 card . another option is a hot swapable hard drive.. in most systems there is a setting in bios to allow a drive to be hot swapped. I forget the name of the setting..

A key thing is not to overload the network , doing so can cause cluster issues here. the cause may be the way our network is set up - something I'm trying to fix. so the cron scripts need to be scheduled so rsyncs do not occur during backups and write to usb does not happen during network transfers etc.

For linux systems turnkey is a great option . check their web site for info on their backup and restore system. it works and the cost to store the backup is inexpensive . Also you've the option to use their software and store the backups using your own storage. There are many templates avail for openvz and lxc . The current templates are for pve 3. Hopefully we'll see some pve4 ones in the next few months.


One question - do you have a nfs server to use for obnam or rsnapshot? I think starting with rsnapshot is easiest.

 

[sahostking]

ok thanks for all the comments really helped.

While I look into this a bit further though ca someone guide me on this. We setup a Proxmox server with 64GB memory and ZFS raid 10 with 6 x 2TB disks.

Now I can install and create openvz containers as we use proxmox 3.4 we don't believe in using the latest version as yet. Rather give a few weeks or few months before moving to newer versions like version 4 so bare with me.

I have a few CentOS NFS servers working and setup in Proxmox nodes. Just set these up last night. Never used them before was so used to HyperV and Data Protection Manager

How do I get it to do live migrations for OpenVZ containers using proxmox vzdump options with nfs?

Getting this error:
INFO: mode failure - unable to detect lvm volume group

We have LVM on NFS server.
/dev/mapper/VG0-backups 3.6T 779G 2.8T 22% /backups

 

[RobFantini]

Hello sahostking : regarding vz migration with nfs, please start a new thread.

 

[term]

Rob-

Yes, we have NFS available, so far we are just using rsync over SSH. The backups are timed so that they don't overlap too much. Backups are stored on a zfs system, so that provides some snapshot capability. rsnapshot looks pretty good - it uses rsync and stashes a few versions away using hard links. Obnam looks a little more complicated.

 

[RobFantini]

for pve systems we include at least these:
/etc/
/etc/pve/
/var/lib/pve-cluster/

there is not too much that changes on a pve system.

from within vm's you'd know how to figure that out. usually /etc , /var and /opt cover most .

I also run rsnapshot locally . having /etc going back 3 months is good to have around.

Also from man page: parameters must be separated by tabs.

 

[vikozo]

maybe Bacula
is a solution?

 

[RobFantini]

maybe Bacula
is a solution?

1-st I am not an expert on all possible backup methods!

the main thing is to use what you know works and be familiar with it.

to quote the warning after a Turnkey backup:
TIP: test your backups with a trial restore BEFORE something bad happens.

 

[SamTzu]

Honestly..? I'm having a problem believing people still use Bacula? It's like BackupPC way too complex to setup and use easily. Someone should have put them out of their misery long ago...
On a side note have you guys seen what can be done with Laravel/Blade/Envoy setup? I'm thinking this could be the future of backups and great many other things regarding scripting.

 

[udo]

Honestly..? I'm having a problem believing people still use Bacula? It's like BackupPC way too complex to setup and use easily. Someone should have put them out of their misery long ago...
On a side note have you guys seen what can be done with Laravel/Blade/Envoy setup? I'm thinking this could be the future of backups and great many other things regarding scripting.

Hi,
we use bareos (fork of bacula) + bacula (on a smaller site) with good experiences (app. 200 clients with round 25TB data each week).
Perhaps it's not so easy to setup, but it's stable and we often recover data without issues.

Udo