Packet loss

[alan01252]

Anyone experience vast amounts of packet loss today. It's causing lots of general slowness/weird behaviour for me.

[Tash]

Hi @alan01252 - has your connection since improved for you?

If you find you're still experiencing issues, please drop us a message on Facebook or Twitter and we can take a closer look, as well as complete checks on your line.

When using either of our social media channels, please follow the guidance included in the link added as this will ensure you're diverted directly to a member of our team and you won't need to repeat your query.

[Anonymous]

The thing is if you look at the packet loss between you and your end target it's 0%!  So you have no packet loss from your target.

You may have a false expectation of what you should see.  The internet switches along your path will prioritise traffic that passes through them.  Pinging an intermediate switch will at best be treated as a lower priority than data that is passing through, and at worst (there is actually a worse case...) the switch may just be set up to ignore ping requests - hence you see false packet loss!

Having said all that I used MTR (though I prefer pathping in windows) to duplicate your results, and my average access to GitHub on VDSL connection was 32 ms - about 1/5 of that you were posting.  Are you by any chance using an SDNS or similar service?

[alan01252]

Thanks Keith,

I tracked this down to a a Wifi extender in my network causing all sorts of weirdness.

After upgrading the firmware and resetting everything sorted itself out..

My pings are now a respectable 30ms too....

Thanks for your reply!

Alan

[alan01252]

I think I just had just got lucky last night. Problem persists this morning even when plugged directly into the router.

Keith I don't suppose you could please provide a screenshot of your mtr? I am curious to see which routes you're connection is using.

Thanks

Alan

[Anonymous]

Okay so using mtr this is what I get:

My traceroute [v0.94]
router (*.*.*.*) -> github.com 2021-06-25T18:45:56+0100
Keys: Help Display mode Restart statistics Order of fields quit
Packets Pings
Host                                              Loss%  Snt  Last   Avg  Best  Wrst  StDev
1. host-212-158-250-39.dslgb.com                   0.0%   50   9.4  10.3   8.0   65.2   8.1
2. 63.130.105.110                                  0.0%   50   9.3   9.3   8.5   17.2   1.3
3. ae5-100-xcr1.man.cw.net                         0.0%   50   9.4   9.8   8.3   38.5   4.2
4. ae26-xcr1.hex.cw.net                            0.0%   50  16.1  16.3  14.8   24.9   1.8
5. ldn-b2-link.ip.twelve99.net                    34.0%   50  16.1  17.1  15.5   31.4   2.7
6. ldn-bb4-link.ip.twelve99.net                   48.0%   50  16.5  16.7  15.4   22.3   1.7
7. prs-bb2-link.ip.twelve99.net                    0.0%   50  29.8  30.1  29.2   42.7   2.0
8. ffm-bb2-link.ip.twelve99.net                    0.0%   50  29.3  29.8  29.1   35.1   0.8
9. ffm-b11-link.ip.twelve99.net                   64.0%   50  29.4  30.9  29.1   52.0   5.3
10. github-ic350972-ffm-b11.ip.twelve99-cust.net   0.0%   50  29.2  31.2  29.0   49.1   4.0
11. (waiting for reply)
12. (waiting for reply)
13. lb-140-82-121-4-fra.github.com                 0.0%   49  29.5  29.7  29.2   30.6   0.3

*manually edited for clarity

[alan01252]

Thank you,

I'm convinced something isn't entirely right, and although I agree mtr can be showing packet loss for numerous reasons, I suspect tcp packets are also getting lost not just icmp. The fact the mtr you posted shows high levels of packet loss for those hops makes me confident it's not just me too.

For general browsing, it's almost unnoticeable but when doing lots of clones, pulls of repositories, docker images, and various VPN connections throughout the day it's enough to cause disruption.

I'm hesitant to go through the usual support channels with a problem like this, it's usually not worth the agrivation.

Thanks again for taking the time to reply and run the mtr! It's very much appreciated.

[Anonymous]

If you are not seeing packet loss from the endpoint location I'd genuinely put this down to data transfers through the intermediate switches getting priority over pings.

I'm on GitHub at least weekly and I'm certainly not seeing any issues - It does all start getting more complex though once you are using Docker with its virtual interfaces and possibly throwing in a VPN too!

As for the usual support channels, they'd probably think you were speaking a different language!