Encrypted hub device with spoke device local/global mismatched

RPi4 (hub) with several encrypted folders shared with three non-encrypted laptops (spokes) running Ubuntu 25.04. All devices are updated regularly and all have ST 2.0.10.

Changing file on laptop A results in similar local/global state for that folder on the RPI4.

Connecting laptop B to the hub updates those changes but leaves earlier version unchanged or undeleted, even though both RPi4 and laptop B show the folder up-to-date, global and local states are mismatched.

The correct file/folder count is that for the device RPi4

Pause/unpause and rescan have no effect.

Unencrypted folders with several hundred files maintain the correct global and local states across all devices.

How about the states on other devices? Are they the same, or are they all different as well? There is the issue https://github.com/syncthing/syncthing/issues/8735 that can cause such inconsistent states, but the actual directory and file numbers on the disk should still match, and synchronisation itself will also continue to work fine.

Thanks @tomasz86 Other devices are also showing up-to-date, but are in fact out of sync on disk.

I think the error occurs when the second spoke device comes on line – that’s pretty weird. The folder in the screenshots is encrypted on the hub.

I have recently changed my setup from mesh to hub and spoke, so this issue may have been disguised in that arrangement.

Here are the property screenshots from file manager for the mismatched folders that Syncthing shows as up-to-date.

All devices are running SSDs. The RPi4 (4GB memory) , database, data folders and OS are all on a USB SSD.

My theory is that changed/added files are being propagated to other spoke devices but deletions are not, leading to increasing numbers of files (local and global) while displaying up-to-date on all devices. This is pretty weird: totally not how Syncthing is designed.

Anyone with a similar layout or encrypted cloud server is likely to have a similar experience.

If a solution is not forthcoming, I’ll change from receive encrypted to normal send/receive on the hub. Although that is not my preference, leaving data on the hub in plain text while spoke devices (laptops) are protected having encrypted partitions. Maybe I should encrypt the data partition on the RPi4.

Experiment: Deleted receive encrypted folder (Documents) on the hub device RPi4, then recreated the folder as send/receive. Then deleted each spoke device folder Documents and reinstalled from from RPi4. Using one of the spoke devices, I deleted and moved several sub-folders and checked that each action was accurately made on other devices. Every move was accurate and completed almost immediately, as expected.

Conclusion: The effective difference was the change to send/receive on the RPi4, eliminating the encryption for that folder on the hub device.

Would anyone care to try and validate this conclusion?

Any help would be appreciated…

OK, no help forthcoming on this so I have LUKS encrypted the Data partition on my RPi4’s SSD and changed all the encrypted folders to send & receive. Now, there is no discrepancy as a result between all devices – local and global match across all devices (three laptops and Android phone). Booting the PRi4 is slightly now complex over SSH, but no big deal, it works with the partition unlocked and mounted.

I’ll report this experience on the bug tracker and move on. Syncthing is a great project, I’ve been using it for years with great success, thanks to the devs.

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