First, thanks for Syncthing, I use it daily as my personal cloud replacement with an always on server. And it is just great!
Recently I noticed large writes to disk on my server even when there should be only very little transferred.
I use it mainly for small documents/files etc. so I was surprised to see larger then 30Gbs of writes to disk (measured with iotop) in 24hours.
The installation runs on proxmox host → ubuntu server guest → docker install. I first suspected problems with discard not working properly, but I did not find anything. Also atime=off + noatime did not change it.
So I enabled fs logs on Syncthing, but it it only shows what is written, not how much. I suspect that one of the shared folders has a outlook .pst file which is quite big, maybe causing those writes?
Even when nothing much is send over the network, due to chunking and hashing, it might write the whole file to disk every time?
Does anybody have suggestions how I could further look into this?
Every synced file is first written as a copy, with blocks transferred or copied locally, then moved in place. So if you have a 5 GB file and change 1 KB, you’ll see 1 KB transferred and 5 GB written.
Yeah not really. 1 KiB is an extreme case, but consider instead if it were a few hundred megs taking quite a while to transfer and write. What if we crash halfway through? Then your file is trash and we don’t know how to recover it.