My setup is the following:
Raspberry (send encrypted)
/ | \
/ | \
Pixel 6a ---- Galaxy Tab Active ---- Blackhole (receive encrypted)
\ /
-------------------------
All devices are running Syncthing 2.0
The Raspberry is a server, and the share contains SQL DB dumps (encrypted with GPG, but it doesn’t matter for this issue). Each hour, a new dump is created, and dumps older than 30 days are deleted (in general, it means when one dump is created, another one is deleted).
The backup folder contains these dumps. With Syncthing, I sync it to three devices (Blackhole being my laptop, Pixel 6a my phone and the Galaxy Tab a tablet which acts as a last resort backup). When I set it up, I wanted to use the Syncthing encrypted mode.
Every device is connected to each other to prevent NAT syncing issues (Galaxy Tab Active can’t connect to any Syncthing relays due to Android 5 lacking Let’s Encrypt and Syncthing-Fork rejecting user certificates, Raspberry is running in Docker so local sync can fail sometimes), which is what I do for all my shares, and worked correctly for this encrypted share too.
I’m using Staggered file versioning with 30d maximum age, because otherwise crashing the server could delete the files on the other devices too, but it shouldn’t have an impact on my issue.
However recently, I got an issue: on Galaxy Tab Active, Syncthing-Fork creashed without me noticing (the Syncthing-Fork interface doesn’t show any useful informations because it’s stuck in “starting” mode due to an old wrapper issue, even when it’s not running).
When I finally restarted Syncthing on the tablet (after more than a month, I think), deleted files were synced again from the tablet as new files even though the tablet is receive-only.
The state is now:
- On Galaxy Tab Active, share is marked as Unexpected items (5 files which weren’t deleted) but no files download seems to be failing. Both Blackhole and Pixel 6a showed out-of-date with 713 encrypted file deletions pending on each.
- On Pixel 6a, share is marked as Up to date, with Blackhole showed out-of-date (120 encrypted files deletion unsynced from Unknown modifying device, and dated at 2009-02-14 00:31:30).
- On Blackhole, share is marked as Up to date, with Pixel 6a showed out-of-date (147 encrypted files deletion pending from Unknown modifying device, and dated at 2009-02-14 00:31:30).
- On Raspberry, share is Up to date, with Pixel 6a showed out-of-date (20 files pending deletion, deleted by Raspberry through the backup script).
File count is 721 files in global state on each device. Local state shows 721 files on all devices except Galaxy Tab Active (which shows 726 files) due to the Unexpected items.
I already had a similar problem a while ago (2 or 3 months ago) in the previous version of this share, with devices trying to sync deleted files, but failing as no device in the cluster had the deleted files anymore. This problem was caused by reconnecting another device of the cluster after 3 month, which was still running Syncthing 1.X. Back then, I though it was just conflicts due to the major release (which caused a full index resync on all of my devices each time I upgraded one), but I now think it could be the same issue.
Last time, I fixed it by recreating the share, but it’s quite long.
I’m considering removing the encryption fully as dumps are already GPG-encrypted,
Is there any trick to fix this kind of issue? Is there any way it could be fixed in Syncthing somehow?