I have 6 devices that i rsync to a central location to back them up. Ive been using ssh as the -e option. Problem is i use public key with passphrases, meaning to backup all six i need to go to each device and run the backup script. Since i typically backup /etc, /home, and /root this means entering sudo and the ssh passphrase 3x for each device.
I would much prefer a script that runs on back storage device that can pull the data from each device without having to use ssh (encryption is not necessary since all traffic is either local or going through a vpn connection).
I could then put this script in root’s crontab or make it a systemd service running as root.
But i dont know how i can remote sync without ssh
From my understanding, the issue is you can’t run them as background script because it is promoting you for the passphrase of the ssh key?
The easiest way to solve this is to use a ssh key that has no passphrase. Yes it’s possible and it won’t prompt you for it. Whenever you create a key, it asks you to enter a passphrase. If you hit enter without entering anything, there’s no passphrase.
But if you just don’t want ssh at all, you can use rsync daemon. Someone else mentioned it here. It’s not as hard as they said, especially if you’re in a local network where you’re fine without encryption.
rsync with open SSH certificates is secure without prompting for any password at all
Use a restricted account with an un-passphrased key is probably by far the easiest way. You could also use rsyncd, but you’ll have to fool with a whole bunch of stuff. The work involved will probably be a superset of just doing a restricted account for the rsync process to use for rsync-over-ssh.
Edit: I had totally missed that the issue was passphrase of the key, not password
They are already using ssh keys…
Oops. I totally missed that. I revised my comment to reflect it.
And they have to use locked keys? In a setup where they don’t need encryption?
Take a look at borgbackup if you want to overhault your backing up in general. It has some neat advantages.
This. Rsync is not backup. Use an app that was actually designed for remote backup.
That being said, they’d probably have the same problem with borg and password-protected keys. 🙂
I am also going to recommend the same solution as @matcha_addict@lemy.lol in this comment: https://lemmy.ml/comment/7998407
You can create a key pair that is specifically just for this kind of backup transaction.
To limit its affects, create a user and group on each of the devices that are highly restricted.
This is actually the most secure solution that doesn’t require an interactive password prompt. The passwordless key only serves this one purpose and has small attack surface.
Basically, you can tell SSH to allow root login on certain devices by setting up a root key pair. You configure SSH on the target device such that when it logs in, the login must run a script or a single command instead of running a shell, this limits what attackers can do if they somehow steal your private keys. You can also keep these private keys in your SSH agent so you only have to enter their passwords once, this will allow you to run remote commands without a password.
I would recommend also exploring the possibility of setting up an Rsync Daemon on each remote device, it keeps an Rsync process running on a remote device and listens for connections from Rsync clients. https://linuxconfig.org/how-to-setup-the-rsync-daemon-on-linux
On an unrelated topic: you might also want to look into using Btrfs and making and transferring snapshots to other devices.
There’s a somewhat more secure way to go about it.
A backup script on the host in question runs periodically as root and makes a local copy of the files owned by the backup user. The central host then makes a backup of the not-root-owned files on the host in question. It adds two extra steps but you don’t have to set up SSH access for the root account.
I have decided to use rrsync to do this.
Im using btrfs on the back drive but using ext4 on the remote devices. Wont the snapshots, if sent to a remote device be the same size as the original data?
I guess the Btrfs snapshopt approach is not possible for your setup since the devices you want to backup are not Btrfs and cannot create snapshots.
Yes, the snapshots will be the size of the whole partition, I had not thought about that problem. I do not know if it is possible to create incremental snapshots with Btrfs.
You can use rrsync to set up a key without any permissions besides rsyncing to one location.
Have you tried Syncthing?
I use it too but i like rsync better for backups
You can run rsyncd as a service on host you wish to back up and connect to that from your central point directly without ssh. Traffic is unencrypted and I wouldn’t trust on that over public network, but you can bind rsyncd to localhost and open a single ssh tunnel for each host (or even write a small script to keep tunnels open automatically) and then just run rsync over that. That’s how I backup my things, just with backuppc in the mix (I’ve got scripts to open/close ssh tunnels at backuppc configuration). VPN tunnels are also an option to encrypt traffic, but depending on your use case that might be a bit overkill.
Or if you’re not tied to rsync you could use something like BorgBackup or other tools which manage the whole jazz for you out of the box.
You can create a second SSH key without passphrase, add that to the authorised keys, and specify to use that key instead, for example using an SSH config file.
Now this is a bit of a security flaw because anyone with access to the key can use it, so you can use rrsync to make sure that key only has access to rsync, so worst case scenario it can destroy your backup, but that’s expected since you want write access to your backup without inputting a password.
Yeah this is what im leaning towards
Ansible is the way for you I guess
That’s just SSH with extra steps
(But I agree)
Or just remove the pass phrase from the key…
I’ve tried it before but i want a situation where i dont need to use the ssh agent. I think i’m gonna go with using rrsync
netcat?
I love this answer because it’s exactly what he’s asking, but absolutely what he shouldn’t do hahahaha.
Anyone wondering, go to a computer and type
nc -lp PORT > file
(Replacing PORT with the port you want to use), now go to a different computer and typenc IP PORT < FILE
(Replacing IP and Port with the IP from the first machine and the PORT you ran on the command there, and FILE with a file you want to copy). Congratulations, you just copied a file from one machine to another without using SSH.You can pipe tar through it too.
Receiver:
nc -lp 12345 | tar xf -
Sender:tar cf - . | nc 192.168.0.123 12345
Also
dd
if you’re moricated to image over the network.I mean, he asked…
What’re the downsides? I’m sure it’s very insecure. Is it faster?
Sends in the clear, no error checking, the
nc
command is promiscuous while its bound to the port. No crypto or compression to slow you down. Just a raw pipe of bytesIts a bad idea, part of the forbidden codex known only to old, irreverent graybeards who know better but don’t care anymore. There are better ways that are both more reliable and better practice.
You might want to look into using passwordless SSH keys within your script (see
ssh -i
) which isn’t the most secure.practice on multiuser systems, but is Okayish in Devops and backups. Add other factors like aggressive allowed hosts settings on the receiver, and rotate the keys regularly.