Indeed – don’t think it will be a hardware issue then
Hi, takes some time to read and understand this longish thread, a picture about your network topology would make it easier for others to read.
Not clear to me: First network component the 4k+ is connected to is a 5-port giga ethernet switch where also a NUC with iperf3 is present and you can connect your laptop to.
If you directly measure with iperf3 between the laptop and Vero 4k+ … and also between the NUC and the Vero, can you reproduce the bandwidth issue at all with these test configurations?
-
If yes, the error is somewhere in this group: I would give the Vero and the laptop a fix IP and direct conncet them with a LAN cable and redo the test. If the direct cable test is successful your switch is faulty. If the cable-only test also shows the issue only 3 things are left: The Vero4k+, the cable and the laptop.
-
If no, the intermittent error is somewhere in the path to the server behind the 5-port switch. But you can bring your laptop to the next level/switch/router component and redo the tests there. Repeat with all components/nodes on the path to your server till the error re-occurs. That should isolate the issue to the last component you connected your laptop to.
You obviously have a customised network setup if you have multiple switches and LAG setup. I appreciate results from your testing so far is suggesting the 4k is at fault but I would be inclined to suspect the network setup or 4k s/w configuration is causing the issues.
Having a quick scan of your logs something seems odd with routing:
Have you got static IP / manual configuration on 4k? If DHCP, your DHCP system is issuing odd commands based on the routes below
Dec 22 16:36:25 vero4k connmand[335]: eth0 {add} route 172.16.11.0 gw 0.0.0.0 scope 253
Dec 22 16:36:25 vero4k connmand[335]: eth0 {add} route 172.16.11.1 gw 0.0.0.0 scope 253
Dec 22 16:36:25 vero4k connmand[335]: eth0 {add} route 0.0.0.0 gw 172.16.11.1 scope 0
Have your tried a different port on the switch the 4k is plugged into?
Your iperf results (even when using USB NIC) don’t look quite right. You still have dips during the 20s even if momentarily
Run iperf locally on the 4k. Start the iperf server on one terminal and in another run
iperf3 -c localhost -P 3 -O 4 -t 120
What are the results?
Using the USB NIC, run iperf in both directions separately from the NAS device with the following commands
iperf3 -c 172.16.11.40 -P 3 -O 4 -t 120
iperf3 -c 172.16.11.40 -P 3 -O 4 -t 120 -R
What are the results?
Yes the 4k+ is configured with a static IP.
I ran the local iperf test, the results are here - http://paste.osmc.tv/ezipipimuc.avrasm
Since yesterday I have reinstalled the OS and that seems to have resolved the speed issues. What is interesting is if I restore the backup I took yesterday before the restore the issue reappears. I am still doing testing and having to manually reconfigure the box but so far things are working as I’d expect. I have played back several of my high bit rate movies with no issues. I did have one issue with a file share that was causing some weirdness but it turns out I’d typed the IP in fstab incorrectly and it was pointing to an interface on FreeNAS on the wrong vLAN and that was breaking things.
I have no idea what in the previous config would have been causing the issues and I want to validate that things continue to work over the next couple days but I am reasonably confident that the reinstall seems to have resolved the problem even if I don’t understand the underlying cause.
Thank you to everyone that offered assistance and advice.
Turns out my confidence was misplaced. After a fresh install the 4k+ will work for a short period of time and then exhibit the same issues. I’ve about given up on it. Currently I’ve put my older 4k back into service with it’s USB NIC and it’s working flawlessly. The only difference between them is the build of OSMC, the older unit still has a build from January of this year. I am planning to try an older build on the 4k+ as a final test but if that also fails then it has to be something specific to the unit itself.
I’m happy to get this back and investigate personally for you if you’d like.
Hey Sam. I’d be happy to send it back. Let me know where to send it and I’ll try and get it posted to you next week.
I would suggest to do the test suggested before sending a box back and forth.
I don’t really see the point. Given the wealth of evidence that other devices work connected to the same local switch, a NUC, an older 4k with a USB NIC, my LG TV using Plex, all that are capable of playing back high bitrate content I am confident that the intermediate hardware is not the issue.
Additionally, testing with iperf from the NUC and the older 4k to both of my FreeNAS servers and the laptop connected to my core switch show them getting full throughput consistently. This traffic also stays in the same vLAN and therefore never crossed my pfSense router, which currently only has a single gigabit NIC to the core switch, eliminating that as a possible bottleneck or problem.
I have plugged the NUC and older 4k into different ports on the local 5 port switch with no change. And in earlier testing I bypassed the local switch and connected the 4k+ directly to my core switch, which is a run of Cat 6A. The core switch is a 48 port Cisco 3560 that is more than capable of handing the throughput.
I normally run a setup that is pretty customized, a fair number of plugins and Aeon Nox Silva. I thought perhaps one of the plugins was causing problems. In my last bit of testing today I reinstalled the 10/2020 build of OSMC (the latest version available in the installer tool) onto the device multiple times (I do find it odd that after the install it wipes the USB drive, meant I had to reflash the thumb drive multiple times). The last time I did a fresh install and nothing but editing fstab for the shares. It worked for a short period and then exhibited the same strange iperf behavior and inability to play almost any content.
While I do run my own internal DNS, that can’t be the cause as all the testing and the share configuration is done with IP addresses.
I am still going to try an older build of OSMC and will do that before I put anything in the mail.