Arni from Webdock: I think we need hourly Local/Remote website backups on top off the server snapshots.

Of course I can do this with plugins, but this should be done on server level.

Arni from Webdock: Do I assume correctly that the disk-level snapshots are made with the ZFS snapshot function?

In that case it really is impossible to export them in any meaningful way.

However, what could be done is, since you have access to every single LXD / KVM via the host, to pipe the filesystem of that respective instance into either a ZIP archive e.g. to upload it to S3 / SFTP or dump the whole filesystem directly to the remote.

That solution would necessitate additional cost to be paid by the user obviously since we can't expect Webdock to pay for the egress bandwidth necessary for that.

One saving straw could be the Bandwidth Alliance made by Cloudflare among others so the traffic to the remote could be either discounted or free, depending on the company hosting the remote, but that would mean that the parties involved would have to let Webdock join that Alliance, and I don't know if that is realistic at the current size (same discussion as we already had regarding the Webserver control panels).

Epsilon PS _ Paul Schiffer: Thank you for this interesting input. You are correct our snapshots are zfs snapshots.

Question for you both Bart Ronsmans and Paul: Whatever functionality we would build here it would probably not make sense to back up the entire filesystem as it's difficult if not near impossible to restore a full Linux system this way. So, we'd need some functionality of users being able to pick which directories to back up, right?

Also, would this not also mean that it would be a good to have pre and post hooks implemented, so you could for example dump your databases to files which would then get rsynced - or whatever - to the remote?

All of this is so "customizable" on a per customer basis and requirements would differ so much that I have a hard time seeing it would make sense for Webdock to implement something like this. There are so many tools for backup management on the file system level already and so many different use cases that the problem space kind of explodes a little bit here.

We would open a pandoras box as users would start asking for GDrive and Dropbox integration on top of S3 and SFTP as well as convenience integrations for mysql, postgresql, mongodb and others, just to mention a few things which popped into my head just now.

I hope you follow my line of thought here. I honestly believe that our whole server snapshotting is really the only thing Webdock should offer here and any file-system backups should be the customer responsibility to configure and set up as they like.

A middle ground could be we provide some guides and maybe "1-click-install" of some backup tools which users can then configure as they like ...

Arni from Webdock: Am I wrong to assume that any file-based backup can fully restore a Linux system? For example this page can be found in the Ubuntu docs: https://help.ubuntu.com/community/BackupYourSystem/TAR

This describes exactly that, backing up and restoring an Ubuntu system via a TAR archive, so if that post isn't absolutely wrong, it should work.

Did you run into issues with this way in the past because you say it is difficult / impossible to achieve a restore this way?

Coming back to the main topic, I absolutely agree with you that opening up a Webdock-specific backup solution to connect to the usual platforms is a huge can of worms and wouldn't be worth working on since many dev hours are needed.

But, there is a way, because of https://webdockio.canny.io/feature-requests/p/object-storage, that a middleway would still achieve the desired results. Once paid S3 storage from Webdock is available, so reachable by the servers via the internal network, you could dump a filesystem backup on a user-created schedule there and then offer a "managed rclone" instance where the user would enter their rclone.conf into the dashboard, which would then, paired with a schedule, upload these backups to any target the user could want. And since rclone has a huge library of integrations already, you wouldn't have to worry about writing integrations for all those targets we users could want.

And, as a side effect, Webdock could start sponsoring rclone because of the offered managed instances, therefore strengthening a very good open source project.

Epsilon PS _ Paul Schiffer: It has been some years since I tried restoring a Linux system from a TAR backup and all I remember is that althought it's supposed to work on paper it did not :D

But things may have changed since then of course :)

If it is as you say it is in your most excellent suggestion here, then this brings this whole idea into the realm of the possible.

I will keep this request under review until we have had a chance to do our own investigation here on how to achieve this based on your ideas here.

Thank you for your feedback as always, much appreciated.!