All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Help diagnosing poor network speed
I'm having issues with the upload speed on a dedicated server I got a while back, and I could really use some help figuring out the issue as the customer support has been unhelpful and I don't understand the situation well enough.
The server is meant to be 150Mb/s with a 1Gb/s burst. Any incoming transfers to the server adhere to this, but outgoing transfers are really slow. Trying to transfer a file off the server results in speeds of around 500KB/s, when I would have expected to see speeds of at least 10 times as much. Occasionally speeds burst to around 1MB/s, but rapidly drop back down, within a second or so. This has been going on for days now and it is making the server unusable for me.
Customer support has suggested running an MTR test in both directions, and an iperf3 test, both of which apparently indicate that everything is okay. They suggested I contact my ISP. However, I have tried transferring a file off the server using 4 completely different providers, 2 residential and 2 cloud providers - always the same slow speed. I can't accept that everything is okay - it very obviously isn't when you try an actual file transfer. I also tried testing this booted into a rescue system, thinking maybe I misconfigured something - no change.
Please help, I'm at a loss. Test results below. Just to clarify, the Hetzner server was used to test, it is not the one with issues.
https://zerobin.net/?7f9cb18474ad25c5#HfxM9JV64W3xJ8p/WywkXUfaRb39aUZ3SNermqziUOU=
Comments
Please share the result of the following bench :
Thank you for the response! Here are the results: https://zerobin.net/?843a04e16da3edca#Hwo2oXJPrczJwSbjR4oJMPEzIOF5gGGJsAykXxBj/kY=
The system is currently running in rescue mode, which is why I assume no file systems were processsed.
I don't see anything wrong with this. It looks like the fairly standard cheap network in online (for the non-guaranteed start line or something.) It's about what I get on online at least, pretty much useful to major european hosts and that's about it
I've certainly not had such a poor experience on the previous servers I've owned with them. I would've expected to at least be able to reach a few MB/s. What the heck am I supposed to do with 500 KB/s? It's taking me hours just to sync my Nextcloud folder.
I've easily gotten 100~300Mbps+ to/from hetzner, ovh, etc, but <5Mbps to my other servers in the US, for the price I pay it's kind of just expected and I multithread aggressively. Do you pay for the guaranteed bandwidth addon at all?
I'm running yabs now to see if it's still the same since I haven't measured in a while.
Your Provider is the best direction to help.
We once had some customer having problems with his dedi connection being slow. Turned out something on the Kernel he used, we have installed other Kernel for him and the server could troughput its Gbit port again.
There could be anything.
As per your benchmark nothing looks suspicious or wrong.
Maybe you should open a support ticket so their staff can make some network adjustments if needed
Based on your CPU model I think you are on the "basic unmetered" network?
This is what I'm getting on the lowest Pro plan:
Some of their machines have odd "basic unmetered" notations and some machines have "Premium" notations, I am not sure how they determine which is which though.
Afraid not.
Yeah, this is why I tried in rescue mode, but it doesn't look like it is my config.
I have, but after seeing the MTR and iperf3 tests, they decided there was no issue. But how can the iperf test show a great connection to my Hetzner server, whereas just wget-ing a file is abysmally slow?
That's correct, I haven't ever used anything but the basic from them I think.
Can somebody here explain why the iperf shows good results whereas simply downloading a file is extremely slow? This is what baffles me most.
Is your iperf to your server directly or to online's iperf server?
They are on differently routed networks with different upstreams.
Both Hetzner server to Online's iperf, and Online server to Hetzner's iperf were fine. But I admit I haven't done server-to-server. I will try that when I get back on the computer.
Edit: Well, I guess that answers that.