Bug 37416

confirming - cloning to a server hosted as OSUOSL is fast, in the range of 1.5 to 2.5 MiB/s, cloning to a server in Europe (Gemany) is way slower, <<200Kib/s

examples with a LO-sub-repo (282MB to fetch)
git clone git://anongit.freedesktop.org/libreoffice/libs-extern-sys takes about
~200s (including time used for Counting, Compressing, etc. - but even ignoring that it is a rate of 1.4MB/s)

Same with a server in Nürnberg/Germany takes an order of magnitude more time (gives an average of less than 200kb/s - and I was lucky this time, if it was like that all the time nobody would complain). So it seems there is a problem with outgoing bandwidth capacity, not with git itself.

I've asked our PDX (who are hosting *.fdo) to take a look, if you could post traceroutes for fast and slow hosts to *.fdo that'd be useful, I imagine.

$ /usr/sbin/traceroute anongit.freedesktop.org
traceroute to anongit.freedesktop.org (131.252.210.176), 30 hops max, 40 byte packets
 1  corv-car1-gw.nero.net (140.211.166.1)  0.145 ms   0.205 ms   0.145 ms
 2  ptck-core1-gw.nero.net (207.98.64.179)  2.808 ms   2.775 ms   2.443 ms
 3  ptck-car1-gw.nero.net (207.98.64.139)  2.317 ms   2.277 ms   2.243 ms
 4  psu-ptck-gw.nero.net (207.98.64.98)  2.308 ms   2.347 ms   2.314 ms
 5  * * *
 6  131.252.5.143  6.895 ms   4.141 ms   4.470 ms
 7  annarchy.freedesktop.org (131.252.210.176)  2.443 ms   2.443 ms   2.443 ms

→ transfer speed (as reported by git) 1.5-2 MiB/s


$ traceroute anongit.freedesktop.org
traceroute to anongit.freedesktop.org (131.252.210.176), 30 hops max, 60 byte packets
 1  static.129.32.40.188.clients.your-server.de (188.40.32.129)  0.701 ms  0.755 ms  0.748 ms
 2  hos-tr4.juniper2.rz10.hetzner.de (213.239.227.225)  0.160 ms  0.188 ms hos-tr3.juniper2.rz10.hetzner.de (213.239.227.193)  0.186 ms
 3  hos-bb1.juniper1.ffm.hetzner.de (213.239.240.224)  4.827 ms hos-bb1.juniper4.ffm.hetzner.de (213.239.240.230)  4.790 ms hos-bb1.juniper1.ffm.hetzner.de (213.239.240.224)  4.817 ms
 4  20gigabitethernet4-3.core1.fra1.he.net (80.81.192.172)  11.012 ms  11.180 ms  10.925 ms
 5  10gigabitethernet1-4.core1.par2.he.net (184.105.213.162)  16.099 ms  15.634 ms  14.981 ms
 6  10gigabitethernet1-1.core1.par2.he.net (184.105.213.90)  15.534 ms 10gigabitethernet7-1.core1.ash1.he.net (184.105.213.93)  97.595 ms 10gigabitethernet1-1.core1.par2.he.net (184.105.213.90)  15.057 ms
 7  10gigabitethernet7-1.core1.ash1.he.net (184.105.213.93)  97.530 ms  96.336 ms  94.007 ms
 8  10gigabitethernet1-2.core1.nyc4.he.net (72.52.92.85)  92.967 ms 10gigabitethernet8-3.core1.chi1.he.net (72.52.92.178)  117.124 ms 10gigabitethernet1-2.core1.nyc4.he.net (72.52.92.85)  96.155 ms
 9  10gigabitethernet3-2.core1.den1.he.net (184.105.213.86)  140.956 ms  139.979 ms 10gigabitethernet8-3.core1.chi1.he.net (72.52.92.178)  116.539 ms
10  10gigabitethernet3-4.core1.sea1.he.net (184.105.213.41)  172.201 ms  173.217 ms  173.171 ms
11  10gigabitethernet1-2.core1.pdx1.he.net (72.52.92.10)  172.163 ms 10gigabitethernet3-4.core1.sea1.he.net (184.105.213.41)  171.862 ms  173.147 ms
12  10gigabitethernet1-2.core1.pdx1.he.net (72.52.92.10)  170.198 ms  168.646 ms *
13  * * *
14  * 131.252.5.143 (131.252.5.143)  200.384 ms *
15  annarchy.freedesktop.org (131.252.210.176)  198.803 ms  204.516 ms 131.252.5.143 (131.252.5.143)  199.237 ms

→ trasfer speed as reported by git: 80-150 KiB/s

Should any of the affected libreoffice hackers come by here: as a stop-gap measure, I've put up tarballs of bare repos for libreoffice here: http://dev-www.libreoffice.org/bundles/ (server is in Europe).

Our hosting provider is looking into this (based on this and other reports). I'll keep the bug posted with updates as I get them.

We have done some network upgrades, and now this seems fast (1-4Mb/s) from the off-campus places that I have easy access to.

Can we confirm that it's now fast for everybody?  (For reasonable values of "fast" and "everybody.")

-- 
David Burns                                    <davburns@cat.pdx.edu>
Systems/Network Manager                        +1 (503) 725-5386
Maseeh College                                 Portland State University

Yes, I can confirm that git clone is now fast for the server in Europe as well. Thanks a lot! (getting around 3MB/s now)

Yup, seems faster for me too.  Thanks a lot David, I'm closing this bug now.