claunia

Hi @weirdcrap, Unfortunately I'm not here to offer a solution. But to share my experience, that is basically, the same problem, performance dropping drastically after some time. However my environment is completely different, to be precise: I do not use UnRAID at all. Both my endpoints are in 600/600 FTTH. I tried NFS4 over UDP and TCP, SMB3 over TCP, SSHFS over TCP, rsync over TCP. All thru Wireguard. iperf3 shows 350Mbps (half of the link speed, but, well, it's ok). Speed starts at about 300Mbps, suddenly gets down to 90Mbps. Amount transferred is irrelevant, seems random. After more than half an hour with no transfer, the speeds go back to ~300. Maximum disk speed on source is 500MBytes/sec, on destination it's 300MBytes/sec. Server is on a VMware ESXi VM, with 4Gb of RAM, 2 cores from an AMD Ryzen 3400. Client is native Xeon E5650 with 48Gb of RAM. Both client and server are running Arch Linux with latest updates. When the speeds gets down there is no CPU usage (less than 5%), memory usage is minimal (megabytes) and network is, well, 90Mbps. When the speeds gets down, immediately using iperf3 shows the slow speeds as well. To me it pretty much seems to be something about Wireguard, or maybe their ChaCha algorithm.