old folder pairing let suddenly crash syncthing - help needed

using 5 folder pairs with about 10.000 Files successfully for month now, using ubuntu 14.4 32-Bit

After a reboot of my laptop with next syncthing start, syncthing is crashing again and again. With try and error I could reduce the problem back to 2 folder pairs, each of both are separated file systems

I did a file system check, all fine

I deleted folder , no change, still crashing

If I pause sync to remote partners, syncthing is working

If I start sync, syncthing crash at once

If I pause this two folder pairs, syncthing working too and sync the other 3 folder pairs perfect with his partner nodes.

I deleted complete syncthing setting on this machine and recreate ALL, include new cert+key.pem, pair this new node with the remote nodes. Just in the moment I accept the folder pairing of the “critical pair”, syncthing crashed immediately

the copies of this problematic files are still synced between other nodes

some month ago I had a similar issue with the same folders and machine. I deleted “index-v0.14.db” and restart syncthing a lot of times, suddenly issue was gone and syncing worked perfect until now. But now this don’t get syncing to work again

don’t know what to do any more on this machine to get syncthing running with this folder pairs again.

You could start off by providing the crash logs, as given what we know now, it’s not much.

you mean the panic-< timestamp >.log ? Are there any critical informations inside I had to delete before?

Depends on your definition of critical. You can remove your device ids, folder ids, folder names, ips etc if you are concerned. The most meaningful bit for us is the line that says panic: and then the stack trace after that.

panic-20170919-201303.log.zip (9.3 KB)

hope this helps to identify

Your syncthing config/database lives on a filesystem that is not posix compliant, which we require as we store a database there.

Strangely it only happens for the temporary databases not the main database. Do you know where your config and database is actually located, is it perhaps a symlink to a different filesystem?

Ubuntu (and syncthing) running in a VM (vmware), configuration lives outside of this vm in a shared folder from host system this works perfect for a year with several clones of this vm. That is the reason, why config/db lives outside the vm, not to clode node ids etc.

only one vm (and always the same vm instance) had this problems

and why this problem occurs with 2 of 5 folder pairs only? All use the same configuration and db location.

It happens with 2 folders out of 5 as they are big enough that index sorting has to spill to disk and does not fit in memory.

You can not clone node ids by keeping the key files outside and symlinking them from a different location.

look like you find the trick +1

moved folder inside the vm, deleted database to have a clean startup

Start syncthing → everything works again :star_struck:

Spent a lot of time in this issue. Maybe a hint or feature request. Could syncthing check at startup, if the place of config/database satisfy his requirements?

thanks again for your support !

It already does, it should not be able to start in your case, so I suspect something is fishy happening with paths so that permanent db and temp dbs end up on a different path.

Also with what you did, you can expect the issue to reappear if you modify a lot of files in one go or while syncthing is off.

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