.syncthing.tmp files not going away

0.14.47 (problem started on 0.14.44 though), two OSX nodes (A and B, where the problem is occurring) and two linux nodes (C and D, no problems on either of those).

The shared folder is generally well-behaved, but I’ve got this persistent problem with .syncthing.tmp files in one subdirectory. I’ve deleted them, I’ve restarted A and B, I’ve reset the data base on A and B… upgraded from 0.14.44 to 0.14.47 and reset the databases again. I cannot get these files to stop being created or better, to just sync properly.

They’re regular text files (C source code) that were created on A and never edited or changed on any of the other nodes.

I’m not out of disk space on any of the nodes. A, B and C are on the same network most of the time, but A is a laptop. C is using versioning storage and D is on a public IP in a data centre.

I have syncthing on A started with model debugging on but there’s a lot of traffic and I’m not entirely sure what I’m looking for. Any assistance in helping me understand why this is happening an dhow to make it stop would be greatly appreciated.

These files hang around for 24 hours if the sync fails, which implies it’s failing and completion should not be 100%. I suggest you check the logs.

I understand why they’re created and that they hang around for 24h, but I’m lost as to what I should be looking for in the logs. The logging facilities for syncthing are great but a little overwhelming for non-developers who must gather the info to help those who can help. :slight_smile:

1 Like

If anything is wrong, you should see pull errors in the logs and failed items in the UI. I suggest you look at the logs in the console and not the web UI.

Just an update – The issue was due to OSX node 1 (laptop) making some directory case changes (Src renamed to src) and not being connected to OSX node 2 (desktop) for a week; the filesystems are not case sensitive, and while the two other nodes (both Linux) correctly propagated things, the desktop OSX node couldn’t manage.

I had to manually rename the directories to something different (e.g “src-x”), wait for the sync to take place, see that the desktop node had re-created the Src directory, erase it, wait for the sync to occur and then finally rename “src-x” back to “src”.

I’m grateful that the syncthing development team takes this conservative approach to syncing. I’d much rather complain about something acting weird but not losing data than end up with a corrupted system or a system with data loss.

Thank you.

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.