need to set up a new backup solution (linux, VMs, offsite)
Posted by pirx242@reddit | linuxadmin | View on Reddit | 12 comments
My current solution is mostly file based backups, spiced with own scripts for backing up complete VMs and shipping the backups offsite. It does what its supposed to, but has many gaps. The whole situation could be much better:)
I have
- a few Linux servers (Debian 12)
- a few ESX hosts (version 8 and 7), containing mostly Debian VMs
- 2 Proxmox hosts, containing mostly Debian VMs
- one Windows server (2019) - doesnt really need to be backed up, only has a few windows-only admin tools installed
- almost all servers above are Dell servers (raid, drac and all that)
What i feel is missing that i would want to achieve is
- possibility to backup and redeploy a whole VM (incremental backups if possible)
- redeploying/installing a whole physical server would be nice too
- having stuff synced offsite (not tape) - incremental/diff style
I would still want to be able to recover single/specific files from X days ago though.
Is there anything that could handle all/most of this? Or at least the "whole VMs" and "syncing offsite".
(Or should i just use something like DRBD for offsite?)
I have glanced at
- bareos - seems nice. no offsite though?
- veeam - (we can pay no problem) had a look at the webpage but it was so full of buzzwords it made me sick (and none the wiser)
MorphiusFaydal@reddit
Veeam can do all of that. It'll happily talk to vCenter for the VMware backups, and it has support for Proxmox now too.
For the bare-metal servers, you can use the Veeam Agent (Debian 12 and WS2019 are both supported)
Veeam supports all your wants as well.
Electronic-Sea-602@reddit
Veeam is always a first choice when a company has a budget. I have CE running at home and it is more than enough for me.
Electronic-Sea-602@reddit
DRBD is almost never a good idea for anything, it just sucks in so many ways! If you wanna push data offsite, there's a few solid options to choose from:
1) rsync + SSH. Dead simple, great for scheduled backups!
https://www.digitalocean.com/community/tutorials/how-to-use-rsync-to-sync-local-and-remote-directories
2) rclone. Killer for cloud stuff (S3, GDrive etc), supports encryption too.
There's https://rclone.org/guide/ to start from and https://rclone.org/docs/ in general.
3) syncthing. Real-time sync, peer-to-peer, good if both ends are always online. Probably an overkill for your use case...
https://docs.syncthing.net/intro/getting-started.html
Got ZFS? ZFS with its send/recv over ssh is super-efficient for snapshot-based replication!
https://openzfs.github.io/openzfs-docs/man/8/zfs-send.8.html
and, a bit outdates but still soild https://pthree.org/2012/12/06/zfs-snapshots-send-and-receive/
Best of luck, buddy!
SuperQue@reddit
These are things that don't realy follow current best practices. You want to have automation in place that allows you to reproduce whole VMs and restores data from backups.
It would be a better use of your time if you worked on automation and data backups. Look into using modern backup tools like Restic or Kopia.
Backup data, automate VM creation and restores.
pirx242@reddit (OP)
Most VMs are unique, it would be unmanageable overhead to automate deployment of each one (as far as i can imagine).
Restic and Kopia look nice and elegant, but seem to be "ordinary" file based backup systems in the end.
SuperQue@reddit
Machines being unique makes it even more important to automate. Tools like Ansible make sure that all of that uniqueness is documented in a way that you can recover from failure.
This is sysadmin 101 stuff.
gmuslera@reddit
What changes in your VMs? If you have very few apps in them (cattle more than pets) there should be few directory trees with changes. And databases, probably.
So:
* Backup VM images when there are big changes. In proxmox you have backups, in esxi you can export to ovf.
* Backup DB dumps frequently (depending on the db, you can have a slave meant for backups)
* For those directory trees, check BorgBackup, that does deduplication so is very efficient having several days or weeks of backups
Try to follow the 3-2-1 pattern. For offsite you can use s3 with intelligent tiering, for VMs will automatically go to glacier, for borg backups the unchanged blocks from old backups will not be modified so they can end eventually in glacier too.
pirx242@reddit (OP)
Well, most VMs have different purposes, and are different. Some basically only have one directory/DB/etc that changes, some regularly get modified everywhere, and yet others almost dont change at all... Isnt it like that in most companies/homelabs/etc? :)
gmuslera@reddit
The only problem for my schema are the ones modified everywhere, but you need to understand why, the system, and if it matters for a backup.
_Buldozzer@reddit
I use Acronis Cyber Protect Cloud as a MSP. Very happy with it. The only annoying thing is the SanpAPI Module you have to reinstall after every kernel update, if you backup physical Linux servers or you have the agent inside the VM.
474Dennis@reddit
For anyone wondering, Acronis SnapAPI allows creating snapshots during the backup like Volume Shadow Copy Service (VSS) in Windows.
And yeah, it must be compiled for each kernel. I just want to mention that it can be pre-compiled like mentioned here and here. I hope this will ease management of Linux agents.
_Buldozzer@reddit
Thank you! This will definitely help!