Some $few entries out of $many need GC messages are expected, not indicative of any corruption.
Ok, so the dbs are correct. Indexes from remotes might still be missing. Anyway coming back to my earlier comment:
Even more so now. Could you post a screenshot of the web UI with the folder that misbehaves expanded, and an explanation which problem is occurring (if it isn’t obvious from the screenshot).
Syncthing uses ~/.config/syncthing, if the db already exists there, XDG_DATA_HOME if the env var is set and xdg-default aka ~/.local/share/syncthing only if that directory exists already. stindex probably cut some corners there - the upcoming syncthing cli index that replaces stindex will use the proper location.
Here it is:
Changed files get synced immediately, but there are thousands of unsynced subfolders, especially evident by the staggering difference in Global State. To my knowledge, the Global State should never differ.
If the global state differs from the local state, one of the participants may have different information about the ignore patterns. If this difference causes the total number of items to be higher than the others, this participant quasi leads the global state. If it makes sense and is possible, the ignore patterns should therefore be the same on every participant device. Otherwise this is not an immediate fault.
If the synchronization remains at a percentage, it could be an permission problem, after which the sync may then run through. If this is not the case, it is possible that a scan process is still running on involved devices and the synchronization could only be finished after the scanning has been completed.
Ignores are unlikely to be the cause for global state differences in general, and almost certainly not in this case, as all devices are connected to each other.
To me this looks like an occurrence of the mystery missing index flaw. Unfortunately the cause isn’t known. Running with the --reset-deltas command line flag once on pavilion-arch should fix it.
Should I expect lots of conflicts when doing that?
Ignore files are fully in sync, so that shouldn’t be an issue.
No that shouldn’t happen. And there was a bug fixed in v1.17.0 that can cause the problems you see here. The equivalent of --reset-deltas happens automatically if you upgrade to v1.17.0.
Ah well, I’m on Arch so I have updated to 1.17 right away 
It seems better now, though the global state isn’t fully in unison yet, I will keep an eye on it.
Currently, pavilion keeps believing in its own local state as being the global one, while the global state as listed on elephant is growing beyond 240 GB.
finally, after waiting a while, they came in sync
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.