While I do agree that the thread is a little confusing, it can be divided into two very different problems, which I’ll try to detail below. If anyone disagrees with what I say, or thinks that I’ve forgotten something important, please feel free to chip in.
I’ll deal with @floatingkiwi’s problem first. This issue seems to be a randomly occurring slowdown of iperf3 itself and, on the basis of evidence presented so far, is not related to a slowdown of the network itself, nor has it been suggested that the performance of video streaming has been degraded. After the iperf3 slowdown has occurred once, all subsequent runs of iperf3 are also slow until the Vero4k+ is rebooted.
iperf3 figures show that, when the problem occurs, data throughput drops from around 945 Mbps to around 100-200 Mbps. In this case, the ethernet interface has clearly not renegotiated to 100 Mbps.
floatingkiwi has already confirmed that the problem occurs even when the Vero4k+ is directly attached to the same router as the PC running the iperf3 server. From this we can exclude it being down to a network issue per se between the media room and the office PC and we can infer that it’s not an issue with a switch, cable or port on the switch.
I have confirmed, via ifstat, that there is nothing else consuming the ethernet bandwidth on the Vero4k+ when a slowdown occurs. If it had been a congestion issue elsewhere on the network, I don’t see how a reboot of the Vero4k+ would clear the problem.
The iperf3 figures (using the -V flag) suggest that iperf3 itself is the problem (or at least a part of it). When the slowdown occurs, iperf3 uses high CPU and the process seems to become CPU-bound. I have examined the iperf3 source code and the CPU figures shown by iperf3 are for the iperf3 process itself, and are not for the whole operating system.
There seems to be some kind of state change on the Vero4K+ that makes iperf3 run in low-throughput / high-CPU mode. I don’t have a Vero4k+, so can’t investigate this further.
One outstanding test that @floatingkiwi can run is to see if thie issue is confined to iperf3 or if the slowdown, once it occurs, affects other network traffic, eg NFS and/or SMB.
@Ainsley’s problem seems simpler. The Vero4k+ is showing a relatively large number of dropped ethernet packets and the interface is renegotiating from 1 Gbps to 100 Mbps. I’ve asked for a before and after ethtool report and full logs after the renegotiation has occurred. We can take it from there.
