Re: Measuring testing vs stable vs oldstable usage from security mirrors logs
On 08/08/13 at 16:45 +0200, Simon Paillard wrote:
> Hi Lucas,
>
> On Thu, Aug 08, 2013 at 11:00:15AM +0200, Lucas Nussbaum wrote:
> > I would be very interested in getting an idea of testing vs stable vs
> > oldstable usage, and I was wondering if the security mirror logs could
> > be used for that.
>
> Note we can access ftp.debian.org backends, that may give different figures
> that access to security archive.
>
> We also have the chance to have some very responsive admins of officials
> mirrors, we can come with a script they may run.
>
> > The idea would be to count the respective number of http hits on e.g.
> > http://security.debian.org/dists/squeeze/updates/main/binary-amd64/Packages.bz2
> > http://security.debian.org/dists/wheezy/updates/main/binary-amd64/Packages.bz2
> > http://security.debian.org/dists/jessie/updates/main/binary-amd64/Packages.bz2
> > (maybe filtering with distinct IPs)
>
> IP address are no longer present in access log, I guess DSA did that change
> for anonymity purpose, but it makes conclusions from logs parsing more difficult..
>
> > As I'm mostly interested in relative usage, the logs don't need to be
> > consolidated between the various mirrors.
> >
> > Do you already have statistics about that?
> > If not, would it be possible to do a one-off run now,
>
> That gives, on August 2013, grepping Packages hits on security mirrors (before
> tests with fastly):
> squeeze or oldstable
> lzma 963925
> bz2 20508309
> gz 3217917
> total 24690151
>
> wheezy or stable
> lzma 27859
> bz2 5594851
> gz 143140
> total 5765850
>
> jessie or testing
> lzma 2908
> bz2 561183
> gz 32010
> total 596101
>
> so:
> squeeze 79.5%
> wheezy 18.5%
> jessie 1,9%
Interesting. Using popcon submissions (methodology described at
http://www.lucas-nussbaum.net/blog/?p=775), I get very different
proportions:
squeeze: 41%
wheezy: 48%
jessie: 11%
Too bad we can't filter by distinct IPs...
Lucas
Reply to: