We are using Syncthing to sync our characters on an ARK cluster. It works great for a few hours after I restart it, then it just starts taking 1-4 minutes to synch character files when a player transfers. Once a player transfers to a new map there information is copied into a temporary folder (Its not a very big file) then the other machine pulls that data and puts in the appropriate place. The service works great for several hours, players are able to transfer with no problem, then out of no where it just starts lagging behind and taking forever to transfer the data. How can we fix this issue?
Please provide screenshots of the Syncthing Web GUI and possibly logfiles, so that others can at least have something concrete to work with. Is it the transfer that is either stuck or moves very slowly, or are the changes simply detected late? More information on the Syncthing versions, OS, hardware, etc. would be useful here.
I’d also suggest to change the title to make it actually descriptive regarding the issue at hand.
I will work on getting screen shots/logs. So these machines are based in the same datacenter but not on a local network with each other. So far what I have tried was changing the options to sync newest files first which helped improve the amount of time the issue is resolved for. Keep in mind the temporary folder is always empty until a file is moved to it to be transferred. So there is not really a lot of files for it to keep track of or anything. I guess the issue is more of a “It is slow to find the file that needs synced” Once the file is found it syncs instantly as both machines are on a 1gbps connection.
Machines are as follows
Ryzen 5950X 128 GB DDR4 3200 RAM two 1TB NVME M.2 SSD 1gbps connection
Ryzen 5600X 64GB DDR4 3200 Ram 2x 1TB NvME SSD 1Gbps connection
Everything works great for several hours, then it just simply seems to forget to sync. Again I will get some logs and screenshots as soon as I can.
Yeah, so this sounds more like a problem with the file watcher being slow to detect the changes rather than syncing itself. The operating system will probably be the main actor here (e.g. there are known issues with FreeBSD).
Also, this may end up being just speculation, but you’ve been talking about using temporary folders, so I will mention this just in case. Basically, the file watcher is bound to a specific Syncthing folder. If you remove the folder physically from the disk, and then recreate it, the watcher will not work even if it’s been recreated under the same path. I had this problem before, and from my memory, a full Syncthing restart was required to make the watcher be able to detect changes in that folder again.
Oh yes I am sorry we are using Windows 2019 Server Standard.
So is there a way to make it just restart periodically?
You can always switch to periodic scanning, but please keep in mind that there is a performance impact associated with it. It shouldn’t matter much on the hardware you listed above though, unless the folders in question are huge and/or located on a spinning drive.
I don’t think there are any known issues regarding the file watcher in Windows, but I’ve personally encountered some delays from time to time, usually when the OS was under very heavy load. I’ve also had a few cases, where the watcher just stopped working silently for no specific reason, which sounds similar to what you’re experiencing. It was very rare though, and I still don’t know what the actual culprit was .
It also depends on how the files are written. Some applications memory map files, which means modifying them does not update the modification timestamps or fire any watch events, making it impossible to detect changes, which might explain why restarting fixes things.
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.