Admin stats
A word of warning: try not to compare VCSs using the stats provided here: first, they are partial (only HTTP), and second, each VCS has its own usage pattern, repository viewers may or may not be hidden from search engines, HTTP may or may not be used as the common transfer protocol, etc. However it's fine to analyse the evolution of one VCS on its own.
- Cacti:
http://savannah.gnu.org:8080/cacti/graph_view.php
- memory, load, CPU, disk I/O, disk temp, bandwidth, ...
- Munin:
http://savannah.gnu.org:8080/munin/
- an simpler (to install) alternative to Cacti
- Awstats (BEWARE: 2008's bandwidth is exagerated because Apache didn't log partial downloads properly; also awstats stores resumed downloads / code 206 as unseen traffic, so the yearly stats are largely underestimated):
- http://savannah.gnu.org/awstats.pl
- http://download.savannah.gnu.org/awstats.pl
- http://audio-video.gnu.org/awstats.pl
- Browser-only (not the VCS traffic itself):
- Webalizer (BEWARE: 2008's bandwidth is exagerated because Apache didn't log partial downloads properly):
- http://savannah.gnu.org/webalizer/
- http://download.savannah.gnu.org/webalizer/
- http://audio-video.gnu.org/webalizer/
- Browser-only (not the VCS traffic itself):
- Mirrors freshness: http://download.savannah.gnu.org/mirmon/
- Necessary for us as well: here's a post from Derek Price (a CVS hacker): "Sylvain, can you give us any statistics on Savannah? Average number of concurrent users, frequency of commits, frequency all users exit, etc.?". It would be good to be able answer such questions.
- Apache: http://savannah.gnu.org/server-status/
- PHP XCache: http://savannah.gnu.org/xcache/
User stats
We have about no stats. It would be good to fix that eventually.
- For users:
- Lots of stats could be done via on-commit script that would log the commit in the database. Drupal has a 'project' module that extensively uses this. Check https://savannah.nongnu.org/projects/drupal-forge/ and http://drupal.org/cvs
- Download areas and webpages: generate stats, or even better try to split the logs per-project (though... more than 2500 open file descriptors is likely to break). Now that we have mirrors this may not be much relevant anyway.