LES from Inception in Dallas: packet loss to/from Europe

ataribasementcluster

I use two VPS on the Dallas node from Inception for proxying. For a few months now I've been experiencing 2-10% packet loss on both, in general, in bouts, and at the connection initiation:

root@mylocalmachine:~# ping -f [we still don't post IPs, right?].98 
PING 56(84) bytes of data.
................^C   
--- ping statistics ---
214 packets transmitted, 198 received, 7,47664% packet loss, time 3361ms
rtt min/avg/max/mdev = 242.600/278.443/320.271/17.202 ms, pipe 21, ipg/ewma 15.780/252.115 ms

it's similar to/from other euro machines in various directions, here's to OVH for example over IPv6, no difference really

root@LESVPS:~# ping -f my.machine.in.ovh
PING 56 data bytes
......................................................................^C 
--- ping statistics ---
521 packets transmitted, 451 received, 13% packet loss, time 8398ms
rtt min/avg/max/mdev = 157.012/160.179/197.734/4.014 ms, pipe 14, ipg/ewma 16.150/159.652 ms

once the connection is over the pond, it's perfect

root@LESVPS:~# ping -f 8.8.8.8
PING 8.8.8.8 (8.8.8.8) 56(84) bytes of data.
.^C 
--- 8.8.8.8 ping statistics ---
752 packets transmitted, 751 received, 0% packet loss, time 10536ms
rtt min/avg/max/mdev = 13.024/14.164/40.957/2.508 ms, pipe 4, ipg/ewma 14.029/16.485 ms

What's going on? Other LES nodes from Inception are fine but, obviously, the US proxy is most useful for getting around GDPR/cookie/geoIP nonsense.

MaxKVM

MTR/traceroute data both to and from the system is usually helpful (or required) to diagnose network/packet loss/latency issues.

Why not open a support ticket with Ant?

InceptionHosting

You kind of gave yourself a clue when writing that no?

@ataribasementcluster said: once the connection is over the pond, it's perfect

It is losing packets somewhere in Europe? probably does not impact TCP either, just ICMP.

Do an MTR you will probably find the likely source.

It should be noted that the IPv6 is just an he.net tunnel.

imok

I didn't know Inception had services in Dallas.

InceptionHosting

@imok said:
I didn't know Inception had services in Dallas.

Oh good point, Dallas was shut down around 4? Years ago and moved to Phoenix.

ataribasementcluster

@MaxKVM said:
Why not open a support ticket with Ant?

The support for LES has always been through forum.lowendspirit.com which now redirects here. Also, it's probably not strictly a problem with Inception.

@AnthonySmith said:
It is losing packets somewhere in Europe? probably does not impact TCP either, just ICMP.

It does impact TCP, my SOCKS proxy is getting disconnected or lags.

Do an MTR you will probably find the likely source.

213.46.178.34 and 62.115.125.128 are showing packet loss, the first one looks like it belongs to my provider (UPC aka Chello aka Liberty Global) and the second one is Telia. So I should complain to my provider?

InceptionHosting

@ataribasementcluster said: 213.46.178.34 and 62.115.125.128 are showing packet loss, the first one looks like it belongs to my provider (UPC aka Chello aka Liberty Global) and the second one is Telia. So I should complain to my provider?

you can, but all they will say is that it is packet deprioritisation due to the volume and blame the number of people working from home because of covid.

ISP's in the UK at least do the following for 'residential' connections

When things are busy, ICMP starts to be dropped, then the start to drop VPN/proxy traffic (they can detect it) because they consider that a commercial requirement.

It sucks but that is what happens in most places, there was a big investigation into it about 8 years ago when it was discovered that plusnet (big UK ISP) were de-prioritising VPN and Proxy traffic on their home plans in same cases restricting it to 10% of the speed you paid for.

You said it impacts the connection between OVH and Phoenix too though so you may have a better route tackling that as its obviously not home-based, but I am not sure if fixing that actually fixes your issue in any way though.

havoc

@AnthonySmith said:

When things are busy, ICMP starts to be dropped, then the start to drop VPN/proxy traffic (they can detect it) because they consider that a commercial requirement.

huh? Half the business world is using VPNs at the moment...

InceptionHosting

@havoc said:

@AnthonySmith said:

When things are busy, ICMP starts to be dropped, then the start to drop VPN/proxy traffic (they can detect it) because they consider that a commercial requirement.

huh? Half the business world is using VPNs at the moment...

Yep, and they keep the commercial customers happy by restricting the residential ones.

havoc

@AnthonySmith said:

Yep, and they keep the commercial customers happy by restricting the residential ones.

No I mean all the suits are all WFH on their residential connection. If they can't connect to office via VPNs and thus can't do their job they are changing their ISP asap so I can't see a residential ISP messing with that. It's a rich demographic who are likely also the breadwinners in the house and thus decision maker over spending. Just doesn't seem plausible that ISPs would intentionally cut VPNs as a priority when things get tight. Torrents & a million other things would be first

InceptionHosting

It’s just a fact of life, they are doing it and have invested fortunes in the ability to accurately detect that type of traffic.

This point in time is fairly unique, maybe they have relaxed it but one thing is for sure, UK isps oversell residential dsl lines worse that cvps oversells openvz containers. They have to slow something down when it gets saturated.

They justified by saying a VPN is a business tool if you want unrestricted use then pay business rates for your DSL line.