apt-get updates proxied at Fremont NOC?
Edit: Here's an example. Slightly mangled due to screen refreshes but you get the point:
Get:4
8 Replies
caker
Linode Staff
Which mirror are you hitting?
-Chris
I just jumped over to a Linode in your UK NOC and the same apt-get (hitting the same mirrror) flew by at 141kB/s the few times I was actually able to see the progress counter.
Edit: Another attribution here so you don't think I'm nuts:
My linode is in Dallas. The same update in my home PC with a 3 Mbps connection took less than a minute.
Here is a traceroute from my linode to security.ubuntu.com:
traceroute to security.ubuntu.com (91.189.92.166), 30 hops max, 60 byte packets
1 a1.7.1243.static.theplanet.com (67.18.7.161) 0.526 ms 0.598 ms 0.608 ms
2 xe-2-0-0.car03.dllstx2.networklayer.com (67.18.7.89) 0.183 ms 0.203 ms 0.184 ms
3 po101.dsr02.dllstx2.networklayer.com (70.87.254.77) 0.458 ms 0.526 ms 0.541 ms
4 te4-4.dsr02.dllstx3.networklayer.com (70.87.255.133) 0.670 ms te7-4.dsr02.dllstx3.networklayer.com (70.87.253.121) 0.753 ms te4-3.dsr02.dllstx3.networklayer.com (70.87.255.129) 0.780 ms
5 e5-2.ibr04.dllstx3.networklayer.com (70.87.253.29) 0.483 ms 0.523 ms 0.593 ms
6 xe-8-1-0.edge4.Dallas3.Level3.net (4.59.32.29) 0.456 ms 0.477 ms 0.470 ms
7 vlan60.csw1.Dallas1.Level3.net (4.69.145.62) 0.881 ms 0.831 ms 0.842 ms
8 ae-63-63.ebr3.Dallas1.Level3.net (4.69.151.134) 1.057 ms 1.022 ms 0.924 ms
9 ae-7-7.ebr3.Atlanta2.Level3.net (4.69.134.22) 20.875 ms 20.655 ms 20.934 ms
10 ae-63-63.csw1.Atlanta2.Level3.net (4.69.148.242) 20.912 ms ae-73-73.csw2.Atlanta2.Level3.net (4.69.148.254) 20.732 ms ae-63-63.csw1.Atlanta2.Level3.net (4.69.148.242) 20.723 ms
11 ae-61-61.ebr1.Atlanta2.Level3.net (4.69.148.233) 21.301 ms 20.304 ms 20.398 ms
12 ae-6-6.ebr1.Washington12.Level3.net (4.69.148.106) 33.214 ms 32.820 ms 42.763 ms
13 ae-1-100.ebr2.Washington12.Level3.net (4.69.143.214) 33.201 ms 32.851 ms 32.850 ms
14 4.69.148.49 (4.69.148.49) 37.421 ms 37.261 ms 37.305 ms
15 ae-43-43.ebr2.London1.Level3.net (4.69.137.73) 105.774 ms 105.451 ms ae-42-42.ebr2.London1.Level3.net (4.69.137.69) 105.504 ms
16 4.69.143.93 (4.69.143.93) 105.975 ms 105.784 ms 107.106 ms
17 ae-3-3.ebr2.London2.Level3.net (4.69.141.190) 110.041 ms 105.763 ms 105.946 ms
18 * * ae-26-52.car2.London2.Level3.net (4.68.117.48) 105.935 ms
19 gi1-0-1.oxygen.canonical.com (195.50.121.2) 108.192 ms 106.957 ms 106.268 ms
20 vlan931.watermelon.canonical.com (91.189.93.145) 119.528 ms 119.060 ms 119.488 ms
21 atemoya.canonical.com (91.189.92.166) 188.049 ms 186.461 ms 187.446 ms
And from my home PC:
traceroute 91.189.92.167 with 64 packetsize
1: dsl-servicio-l200.uninet.net.mx (200.38.193.226) 12 ms
2: bb-la-grand-5-pos0-15-0-0.uninet.net.mx (201.125.48.82) 62 ms
3: xe-10-0-0.bar1.Tustin1.Level3.net (4.53.180.5) 59 ms
4: ae-0-11.bar2.Tustin1.Level3.net (4.69.136.198) 61 ms
5: ae-10-10.ebr1.LosAngeles1.Level3.net (4.69.136.206) 59 ms
6: ae-2-2.ebr3.SanJose1.Level3.net (4.69.132.9) 77 ms
7: ae-83-83.csw3.SanJose1.Level3.net (4.69.134.234) 82 ms
8: ae-94-94.ebr4.SanJose1.Level3.net (4.69.134.253) 125 ms
9: ae-2-2.ebr2.NewYork1.Level3.net (4.69.135.186) 132 ms
10: ae-72-72.csw2.NewYork1.Level3.net (4.69.148.38) 131 ms
11: ae-81-81.ebr1.NewYork1.Level3.net (4.69.134.73) 131 ms
12: ae-43-43.ebr2.London1.Level3.net (4.69.137.73) 239 ms
13: 4.69.143.89 208 ms
14: ae-3-3.ebr2.London2.Level3.net (4.69.141.190) 203 ms
15: ae-26-56.car2.London2.Level3.net (4.68.117.176) 197 ms
16: gi1-0-1.oxygen.canonical.com (195.50.121.2) 199 ms
17: vlan931.watermelon.canonical.com (91.189.93.145) 197 ms
18: bignay.canonical.com (91.189.92.167) 230 ms
In /etc/aptitude/sources.list I mean.
If you are interested in a local caching proxy, I currently run one in Newark, and hope to add a few more in other DC's in the near future.
@JshWright:
The default "us" Ubuntu mirror is in the UK. I would still expect better speeds than you're seeing, but you should probably try a different (closer) mirror. There's a whole lot of internet between your node and London (as your traceroute shows).
If you are interested in a local caching proxy, I currently run one in Newark, and hope to add a few more in other DC's in the near future.
http://linsides.com/services/aptcache/
I guess what I'm saying is that it seems counterintuitive that a high quality DC like HE would show such a glaring throughput differential compared to my office and house, both of which are here on the left edge of California with a whole lotta internet between here and wherever the default Ubuntu apt cache is. FWIW I hit both the 91.189.92.166 and 91.189.92.167 addresses when trying to get past the even-worse timeouts I was suffering the other night with this.
I know there are myriad workarounds, heck I was even prepared to proxy myself right through NJ, ATL or DAL where there were no issues with my apt-get maneuvers nor in my scp tests between this host and my Linodes in those DCs, but the bottom line is that the whole point of apt is to just make it happen… right? It's not like I'm expecting a default config to perform well in some far-off place, but rather expecting a default North American config to perform reasonably well in a well-respected and performant North American datacenter.
And BTW this is not a gripe, rather a puzzling anomaly that I can reproduce in my Linodes deployed at HE from a couple years ago through ones deployed two days ago.
@JshWright:
Yeah, but the default North American mirror is in Europe…
I think we're in a loop here. Let me rephrase:
The default mirror works great everywhere except Fremont. This seems anomalous, since HE is a big, popular and well-connected DC.