Should I backup my file server VM or just the files?



  • So, watching @scottalanmiller video here got me thinking whether I should be backing up my entire VM file server or should I instead be focussing on just backing up the files?

    It's not that the Windows OS is taking up a ton of space. After all, the entire VM with files is close to 30TB so the OS is simply a drop in the bucket. But am I gaining anything by backing up the entire VM?

    I'm using Veeam to do the backups and to be honest, Scott makes a good point that if the OS ever gets corrupted in some way, I'm going to have to rebuild the server and want to only restore the files to a fresh OS. Restoring from a previous state is very useful if some patch broke something (rare) or if something else goes fubar and I lose a perfectly good VM, I can simple restore the whole thing, but in the grand scheme of things, would I be better off just looking to backup my files rather than the VM? Veeam gives me the ability to restore files rather than the whole VM so, do I gain or lose anything by backing up the whole VM? If I focus on saving only the files, then I'm not stuck with the same operating system to run as a file server.

    Curious what people think.



  • @nashbrydges said in Should I backup my file server VM or just the files?:

    So, watching @scottalanmiller video here got me thinking whether I should be backing up my entire VM file server or should I instead be focussing on just backing up the files?

    It's not that the Windows OS is taking up a ton of space. After all, the entire VM with files is close to 30TB so the OS is simply a drop in the bucket. But am I gaining anything by backing up the entire VM?

    I'm using Veeam to do the backups and to be honest, Scott makes a good point that if the OS ever gets corrupted in some way, I'm going to have to rebuild the server and want to only restore the files to a fresh OS. Restoring from a previous state is very useful if some patch broke something (rare) or if something else goes fubar and I lose a perfectly good VM, I can simple restore the whole thing, but in the grand scheme of things, would I be better off just looking to backup my files rather than the VM? Veeam gives me the ability to restore files rather than the whole VM so, do I gain or lose anything by backing up the whole VM? If I focus on saving only the files, then I'm not stuck with the same operating system to run as a file server.

    Curious what people think.

    My file servers have their OSs on their own .VHDX (virtual disks). All data are on separate .VHDX (virtual disks). I back up the VM as a whole, as well as the File Share registry data.

    This way I can quickly load up a new VM from sysprep'd template, attach the data .VHDX disks, then restore the file share registry data / backup to quickly have the entire thing back up and running as if nothing happened.



  • You will almost never want to restore the files without the OS. That is a DR scenario in general.

    If the server (the OS) craps, you just restore everything with the last Veeam backup. That is the entire point of using a VM and a VM backup.

    If that last backup was corrupt or something, then you are in a DR scenario where oyu may just restore data.

    Scott likes to talk like everyone uses state systems when that is far from the case.



  • The more I think about it, the more I think that backing up the entire VM makes the most since in that it's the simplest approach for the backup. Sticking to only files would likely make the backup process/management more complex (maybe, maybe not) but with a VM backup, I have the option of a full VM restore OR just the files anyway.



  • The only time I wouldn’t backup the OS is if I could recreate the configuration easily (Salt Stack, Ansible, etc)



  • I would lean towards taking the entire VM and backing it up, and if the backup appliance supports it. Mounting of my backups to restore piece meal files if this is required.

    Block level backups are just to simple to not do, if you aren't doing state systems. Which how many people really are?



  • Obviously having some flexibility in what you can restore would be best, so you don't have to go and restore everything, just to get a file from the day before.

    KVM has the ability to mount the virtual disk and restore files, and so does Xen Orchestra (and XS).



  • I think you can do the same thin in ESXi and Hyper-V as well I'm just not sure of how it operates there. Likely just mounting your vhd backups to a working VM and then dropping the files.

    But this goes into a different backup approach. Backing up the data as a block device, rather than the entire VM.



  • @nashbrydges said in Should I backup my file server VM or just the files?:

    The more I think about it, the more I think that backing up the entire VM makes the most since in that it's the simplest approach for the backup.

    Always back up the whole VM from the hypervisor. I always do the whole VM.

    On a file server VM, you should also run a script that runs every few days depending on your environment that backs up the File Shares (Not the files, I mean the registry keys that contain the list of shares and all that). This if you can't restore the OS as it's corrupt in your backups or something, you can do how I said in my first post.



  • I only do data volumes but I’m not running any Windows.



  • @nashbrydges

    Currently I do files, even though there are good recommendations and comments regarding the VM level.

    The reason I do files cause I can easily create it and know all the aspects about it and easily test the integrity of the backup via applications like p7zip and calculate things and track it, and things are easier to understand like it is expected for the files to either grow or stagnate.

    On VM level unless you have approach to test the backups like systematic way of getting that VM export and testing it, who knows then if you will restore from it properly when Disaster arrives and if all your VM backups are un restorable then you dont really have a backup. Do note i am not saying Veeam is bad product or not trustworthy, actually I heard the opposite all the time.

    I dont have an automated system for testing exported VM, but this thread got me thinking of virtual-box headless that can be useful for this.



  • @jaredbusch said in Should I backup my file server VM or just the files?:

    Scott likes to talk like everyone uses state systems when that is far from the case.

    Scott likes to talk like everyone should use state systems, which is nearly always the case.

    But even if you don't, if you have automated rebuild and install scripts, or use cloud computing, you rarely want to restore the whole thing rather than just the data. It's way more than just state systems. We've been doing this since long before state systems were a thing.



  • @nashbrydges said in Should I backup my file server VM or just the files?:

    The more I think about it, the more I think that backing up the entire VM makes the most since in that it's the simplest approach for the backup. Sticking to only files would likely make the backup process/management more complex (maybe, maybe not) but with a VM backup, I have the option of a full VM restore OR just the files anyway.

    Something worth noting about this is that if you are using and sticking to only specifically supported apps by your backup vendor and integrating them with agents in some way (like Veeam can do) then this works decently well. If you start using databases or other stateful systems that the backup software isn't specifically programmed to handle, then you either need to create local backups first that the backup software can back up or you need to stop those databases until the backups are done, or you risk that the backups are corrupt.



  • I dislike hypervisor level backups because of the risks and complexities that they bring. You have to be confident that you know absolutely everything that they are backing up and know that all of it is supported. I feel that it is easier to do the more reliable "data only" method, or do both of course.



  • @scottalanmiller said in Should I backup my file server VM or just the files?:

    @jaredbusch said in Should I backup my file server VM or just the files?:

    Scott likes to talk like everyone uses state systems when that is far from the case.

    Scott likes to talk like everyone should use state systems, which is nearly always the case.

    But even if you don't, if you have automated rebuild and install scripts, or use cloud computing, you rarely want to restore the whole thing rather than just the data. It's way more than just state systems. We've been doing this since long before state systems were a thing.

    With a stand-alone windows file server is not so bad redeploying it with the help of powershell. But it’s so different compare to Linux.



  • @black3dynamite said in Should I backup my file server VM or just the files?:

    @scottalanmiller said in Should I backup my file server VM or just the files?:

    @jaredbusch said in Should I backup my file server VM or just the files?:

    Scott likes to talk like everyone uses state systems when that is far from the case.

    Scott likes to talk like everyone should use state systems, which is nearly always the case.

    But even if you don't, if you have automated rebuild and install scripts, or use cloud computing, you rarely want to restore the whole thing rather than just the data. It's way more than just state systems. We've been doing this since long before state systems were a thing.

    With a stand-alone windows file server is not so bad redeploying it with the help of powershell. But it’s so different compare to Linux.

    Yes, much more work for sure. But no matter what OS you use, no matter what your backup situation is, you should have a rapid new deployment system to rebuild systems whether it is script, state, or whatever.


Log in to reply