Speed Dips on Polis Server

[ScanFarer 12]

One thing I have noticed with the Polis server is that there are somewhat consistent speed dips that happen for me every 20-25 seconds, and they last for at least 1 second. It's definitely a first-world problem, but it occurs when I am uploading at gigabit speeds, and the speed drops by at least 500Mbps. I haven't noticed such an occurrence on other servers I've tried. However, I still use Polis because it provides the best average upload speed for me. It really isn't an issue at all for me, though; I just wanted to mention it and see if this is the case for others using the server or if @Staff sees that this is occurring. It might just be some routing bottleneck, but I don't know for sure. I feel kind of dumb posting this because, like I said, it is such a first-world problem.



 

[Hypertext1071 4]

Hi there, inferring from your other post are you using this on TrueNAS scale? I've also noticed that speed drips periodically when using TrueNAS scale (Sadachbia server tho), I fixed it by changing caching settings in my torrent client (qBitorrent). I set disk cache size = 0, I believe that setting that interferes with ARC. After I changed that and did a quick reboot I could sustain near whatever my max speed was at the time and would not notice any of these periodic speed dips. This is just what I encountered tho, I'm not sure if this is related to yours but I hope it helps.

[ScanFarer 12]

13 hours ago, Hypertext1071 said:

Hi there, inferring from your other post are you using this on TrueNAS scale? I've also noticed that speed drips periodically when using TrueNAS scale (Sadachbia server tho), I fixed it by changing caching settings in my torrent client (qBitorrent). I set disk cache size = 0, I believe that setting that interferes with ARC. After I changed that and did a quick reboot I could sustain near whatever my max speed was at the time and would not notice any of these periodic speed dips. This is just what I encountered tho, I'm not sure if this is related to yours but I hope it helps.

Unfortunately, I still notice dips in speed after trying what you mentioned. I am using TrueNAS Scale, but I have also encountered these speed drops while testing with the same server on two other devices with different operating systems. So, I still believe it might be a routing issue or a problem with the server itself, as I have not encountered this with other AirVPN servers. Thank you for attempting to help though.

[Hypertext1071 4]

I see, then it makes sense to think that it's a server issue. In that case, could I ask what settings you use to sustain near 1Gbps speeds? For TrueNAS scale and the application you use. TrueNAS craps on me after the ISP drops the connection and after reconnecting I've found that rebooting the system fixes it, however I would prefer not to do this. So was wondering if there was something wrong with my settings for TrueNAS and the application or is it just my ISP's problem.

Hope you can find a fix!

[ScanFarer 12]

I don't have anything too special for my settings. Set to TCP only, global max connections: 2000, connections per torrent: 1000, global upload slots: 2000, upload slots per torrent: 1000. I'm sure that is not the optimal way to do it, I just haven't tinkered with the numbers since I set it. And I get full gigabit seeding speeds on leech hungry torrents. Not sure about your connection dropping issue though.

[Hypertext1071 4]

It's been a while but a I was also testing on Polis and noticed this as well. Does this still occur for you? I also only notice this on Polis only.

[ScanFarer 12]

On 12/12/2024 at 1:07 PM, Hypertext1071 said:

It's been a while but a I was also testing on Polis and noticed this as well. Does this still occur for you? I also only notice this on Polis only.

Yes, I still notice dips in speed, and it only occurs on the Polis server. It is still consistent with the issue described in my post.

I happened to notice this response from staff in my email, though I suppose it has since been removed.
"Hello!

We have found and fixed the possible cause of the problem (nf_conntrack_max too small for the users' demands). Please test again, you should not see anymore the issue.

Kind regards"

I don’t see any improvements, though; I suppose that could've been why they removed their comment.

[Staff 10017]

2 hours ago, ScanFarer said:

I don’t see any improvements, though; I suppose that could've been why they removed their comment. 

Hello!

Yes, the server with that problem was not Polis, we apologize. On Polis we still can't find this problem, if we could catch it exactly when it happens we could do more, probably.

Kind regards
 

[Hypertext1071 4]

It seems to happen a lot (only a few times when it hasn't) and will continue as long as you are connected, would it be helpful if I could provide a graph when connected to Polis through a support ticket? Hopefully that will help troubleshoot this issue if there is input from multiple users. Thanks!