Server-side usage statistics

[karyon]

It would be nice to have any non-anecdotal data on how the site is used. A solid start would be page views per view, a second thing could be page load times (server-side aka be how long the view takes to be processed)

• a simple middleware that writes to the database like this might be too limiting.
• there is server-side google analytics but that's still google analytics and not quite my definition of server-side.
• i found server-side piwik which seems like a nice idea. it's a rather primitive middleware, but it sends its data to a piwik instance which opens lots of possibilities. if using that, we should be ready to fork and adapt it.

[karyon]

another approach would be goaccess, which analyzes apache's access.log

using it was as simple as apt-get install goaccess and then goaccess access.log --log-format="%h %^[%d:%t %^] \"%r\" %s %b \"%R\" \"%u\"" --date-format=%d/%b/%Y --time-format=%H:%M:%S -o --all-static-files -o /vagrant/report.html

one can add a %D to the apache logging config (and to goaccess' format string), and the time to process each request will be logged and reported. (however, if i understood correctly, this is the time to serve a request (including all the network stuff, so this also depends on the user's connection), not the time django needs to process the request. but it's something.

[tiny-fish-T]

I tried to set up server side piwik from https://gitlab.com/burke-software/django-server-side-piwik but it is very poorly documented, which makes it impossible to set up. Also piwik depends on a fully set up version of Celery, a data broker application which also has very very bad documentation. Seems like server side piwik is out of the equation.

[felixrindt]

I tried go_access and liked it. It's probably an easy solution for the beginning, but it would still be nice to have stats on view performance (esp. because go_access doesnt distinguish views, but paths afaik)