Full file backup constantly starts if multiple servers are attached

client
1

I have 2 servers.
both on arm, one on rasperry pi os and one on Manjaro arm.
Both use btrfs on backup storage.

I connected a few clients to them.

They have different paths to backup, ie, the Manjaro server does the same directories MINUS a few ones.

The client I have on my pc shows this behaviour if it has been turned off and powered on during a time where a backup should have been done.
It starts a FULL FILE (but calls it an incremental on the activity web page) on BOTH servers, starting to transfer 1TB when a full on Manjaro would be only 25GB (with the directories I have selected). And a full is NOT what should run, only an incremental with I would guess 100MB.

If I stop them both, and then starts an incremental one server at a time, it “resumes” the backup but with the correct amount, about 100MB on Manjaro and about 300MB on raspberry os.

I have a period of 1 minute before starting backups, I have changed that to 10min on the manjaro server hopefully not “colliding” the backups.

What can I do to remedy this?
If I don’t sit and actively watch the servers when I start up my computer 2TB (1TB to each server) would be transferred taking HOURS!

Extra info:
Yesterday, I actually let the full file continue on my raspberry os server, took a few hours but the registered data uploaded was STILL only 12,4GB (I’m guessing this is all btrfs metadata or something), because of shared_hashes but it still UPLOADED 1TB and it took a few hours.
How do I stop this behavior?

The clients that run 24/7 does not show this behavior.

2

Is there a way to turn automatic client detection off in the servers?

My understanding is that “the usual arrangement” within any one LAN is that there is one, and only one, server regardless of how many clients and that the server will attempt to associate every client it can find on the LAN automatically. Or maybe it’s the reverse, that the server announces its presence and the client seeks it out. But there does not seem to be any discretion on the part of either end other than trying to find it’s complement to connect to.

I’ve not seen 2 servers on one LAN before, and that might trigger all sorts of status confusion in the clients themselves as to what has occurred when, and with whom.

3

Well, I followed the advice given in the urbackup documentation, adding the key to server_idents.txt on the clients I wanted to connect.

I think you missunderstand. I WANT to backup to both servers, just smaller selection of directories on one of them.

If you are asking, yes, that is more like it. So if a client has already connected and synced to one server, you have to manually instruct it to accept another server.
All according to the documentation.

So I guess I’m asking, is the behavior in the original post a bug or is there some extra consideration I have to take into aspect when setting up multiple servers that is not mentioned in the documentation. (or that I am missing)

4

Bedna,

You could turn off one of the servers, go a day, and see what happens.
It may not be due to having two servers.
Then you could try reversing which server is on and go another day.
Checking “Do not do image backups:” and “Do not do file backups:” on the Settings → Server page will effectively turn off that server.

All for now

5

Thank you for the feedback.

I agree, all of those tests should be made, and I HAVE, hence this second post in the “client” forum. I’m not sure it’s the server but rather the CLIENT that is messing this up, or both.

I DID disable all backups on the second server first, by doing what you say AND remove all directories to be backed up on that server. Restarted server and client. The client STILL tried to start backups to the “disabled” server.

I then removed the client completely from the server (remove client), removed the key in server_idents.txt as well and restarted the server and shut down my client and waited through the cleanup window (until the day after) so the server would remove the client. Started up the client, and the client uploads +1TB to the “old” server when it should be a few hundred megabytes.

I then removed the second server completely by pulling the plug.
I removed the incorrect backups on the “old” server.
I realized ONLY the directories set on the SECOND server was being backed up to the “old server” in the 1TB upload and all previous backups that has been made while the second server has been alive.
There were directories missing so I removed the backups, they were INCORRECT (except for the last small one, I had to wait until next backup before I can remove the latest).
It STILL uploaded 1TB AGAIN!!!

DISABLED serverside (and has always been): Do not do image backups (checkbox filled), Allow client-side changing of the directories to backup, Allow client-side starting of full file backups, Allow client-side starting of full image backups, Allow client-side starting of incremental image backups, encryption and compression is disabled.

ENABLED serverside: Allow client-side starting of incremental file backups, Allow client-side viewing of backup logs, Allow client-side pausing of backups, Allow client-side changing of settings, Allow clients to quit the tray icon, Allow clients to start file restores, Allow clients to configure components to backup, Allow clients to start component restores.
(I don’t have a gui client, but these are the settings that has ALWAYS been like this for I think at least 8 months while the “old” server worked perfectly)

I have NEVER used urbackupbackendctl or made ANY changes on the client, except for the “server-confirms” setting in /etc/default/urbackupclient letting me restore from the server web gui.
NO CHANGES has ever been made to the clients, EVERYTHING is set server side. :frowning:

I can’t trust my backups here. :frowning: :frowning: :frowning: