Using SYD datacentre from Aus - why would packets go via HK?

Questions / Help
1

I have a small test web app - I chose SYD as the region for my own convenience (there are no users) as I’m Aus-based. Oddly (to me), packets go out via HK:

Host                                                                                                                                  Loss%   Snt   Last   Avg  Best  Wrst StDev
 1. 172.28.32.1                                                                                                                         0.0%   208    0.7   0.3   0.1   0.9   0.2
 2. 192.168.0.1                                                                                                                         2.4%   208    1.5   2.3   1.3  37.4   3.6
 3. 10.111.5.188                                                                                                                        0.5%   207   53.7  39.8  23.6 286.0  21.6
 4. (waiting for reply)
 5. 210.49.119.206                                                                                                                      0.5%   207   56.6  39.4  24.5 164.9  15.0
 6. 10.194.254.2                                                                                                                        2.4%   207   57.6  38.0  23.1 107.6  12.4
 7. 10.194.121.213                                                                                                                      0.0%   207   40.0  36.9  23.4  95.3   9.7
 8. (waiting for reply)
 9. bla1-hu0-4-0.ig.optusnet.com.au                                                                                                     0.0%   207   51.8  36.5  24.1  74.7   9.3
10. 203.208.190.109                                                                                                                     0.0%   207  152.1 151.6 137.3 193.1   9.6
11. 63-218-211-253.static.pccwglobal.net                                                                                                0.5%   207  154.6 150.6 137.9 188.6   7.8
12. HundredGE0-7-0-2.br03.hkg12.pccwbtn.net                                                                                             0.5%   207  166.1 151.5 139.2 187.7   8.4
13. (waiting for reply)
14. (waiting for reply)
15. (waiting for reply)
16. 213.188.208.158

No wonder latency seemed a bit high. I’m on a not-so-hot connection in a rural area, but it’s usually 100ms better than this for local content! Also I can see now why I was hitting HK in the Region latency test

I confess to not knowing much about BGP, Anycast etc, so I can’t even make a stab at guessing whether this is an ISP or Fly.io issue (or neither). But it does make my choice of SYD a bit moot.

It’s not of practical concern to me at this stage, but I’d be interested to find out more. If it’s just something that should be obvious based on Fly.io’s architecture and there’s something in the docs that explains this, a pointer to reading material would be great. Cheers.

2

This is basically a routing issue. If you can share your IP address we can likely get it corrected (DM is fine).

3

Thanks - currently 49.195.1.223, but it’s not a fixed ip

4

Hmm … just checked again and it’s fixed. Presumably something you did with miraculous speed?

5

BGP routing is really weird sometimes. Usually we can do something about it with a traceroute and a source IP address.

We did make routing changes that might’ve affected your particular issue!

6

It’s definitely continued to do the trick in my case anyway, so thanks.

Also a useful reminder for me of the value of having a closer edge available. Those reduced latencies really do make a practical difference.

1 Like