Hi, The gap in reports is down to the way Amazon have implemented SimpleDB searching. Despite the search query sent via the Metabase code explicitly stating the search criteria, Amazon's SimpleDB just uses this as a guide and does not necessarily return the result set as requested. As such I have noted that there are many gaps in the data, and I intend to run queries to repull these missing reports. However, what with other CPAN Testers issues, YAPC::Europe, YAPC Conference Surveys, holiday as well as $work, I haven't had time to run these additional queries as yet.
The delay in getting reports from the feed was partly due to the way the feed is processed, and the fact that the database was getting overloaded with queries. The database was reconfigured and various performance improvements have been implement, which has seen the feed now roughly 2 hrs behind and the page builder about 24hrs behind the oldest request. This should remain pretty stable now, although I plan to implement further improvements over the coming months. Cheers, Barbie. On Sun, Aug 21, 2011 at 04:12:09PM -0400, chm wrote: > I still see the gap in the PASS Summary but the good > news is that the 2+ day delay in processing the CPAN > Testers reports seems to have been resolved. > > What ended up being the problem? > > --Chris > > On 8/14/2011 9:45 AM, chm wrote: > >Well, the PASS Summary "glitch" has now spread to > >the PDL data as well for which the missing reports > >are the ones since I took over as release manager > >for PDL. It looks like we're back to the original > >hypothesis that it is related to who releases a > >module. > > > >Also, the CPAN Testers reports still seem to be > >2 days behind. Is this understood or is it > >possibly the result of some internet attack on > >the metabase process? > > > >--Chris > > > >On 8/5/2011 6:01 PM, Chris Marshall wrote: > >>I think I've discovered the problem/feature. The missing > >>reports are for module versions that have been deleted > >>from CPAN to remain only on backpan. > >> > >>I would argue that the PASS matrix should not change > >>just because a module is subsequently removed from > >>CPAN as that is still the most recent module version > >>with a PASS. > >> > >>The use of a CPAN/backpan filter for the PASS matrix > >>could give one a misleading impression of what version > >>of a module is good. > >> > >>Cheers, > >>Chris > >> > >> > >>On Wed, Aug 3, 2011 at 4:25 PM, Christian Walde<mitha...@yahoo.de> wrote: > >> > >>>On Sat, 30 Jul 2011 12:58:08 +0200, chm<devel.chm...@gmail.com> wrote: > >>> > >>>>The CPAN PASS Matrix for OpenGL is still > >>>>inconsistent with the actual information. > >>>> > >>>>It appears that in addition to the builder > >>>>being 2+ days behind that the previously > >>>>determined information is not maintained. > >>>> > >>>>For example, http://pass.cpantesters.org/distro/O/OpenGL.html > >>>>only release 0.57 is reported. I just now > >>>>realized that the missing reports are all > >>>>since I took over releases for OpenGL. > >>>> > >>>>Does that suggest what might be going on > >>>>with the update inconsistency? > >>> > >>>The actual cpan matrix looks fine: > >>>http://matrix.cpantesters.org/?dist=OpenGL%200.66 > >>> > >>>So this seems to be something specifically only for the pass > >>>generator, not > >>>the reports themselves. > >>> > >>>-- > >>>With regards, > >>>Christian Walde > -- Birmingham Perl Mongers <http://birmingham.pm.org> Memoirs Of A Roadie <http://barbie.missbarbell.co.uk> CPAN Testers Blog <http://blog.cpantesters.org> YAPC Conference Surveys <http://yapc-surveys.org> Ark Appreciation Pages <http://ark.eology.org>
