On 16 July 2010 10:45, Maxime Petazzoni <[email protected]> wrote:
> Hi all, > > I've given some more work to my query optimization patch in the past few > days, in particular to track down these annoying GEOS intersect errors. > > It appears the version of GEOS on the server (3.0.x) is very buggy with > intersections computations so there is no real solution here > unfortunately. > > Possible courses of action are: > > * upgrade the server to 10.4 to get Postgres 8.4 and GEOS 3.2, with > which we can use this patch and benefit from the significant > performance improvement. I have tested it on my machine and so far > no rendering has failed with GEOS errors, not to mention the way, > way faster queries. > > I don't know Ubuntu enough to handle this upgrade though. Not sure > of the effect of the upgrade on our infrastructure, too. > > * migrate to the new server as soon as possible, we're waiting to here > from Jerome here on the hardware issues of the machine hosting our > VM. It runs Debian testing, and has Postgres 8.4 and GEOS 3.2. > > * do nothing, and be sad. :) > > Comments, questions and feedback on the patch itself is also appreciated > of course! > > I would advise not to upgrade to 8.4 / postgis 1.4. Frederick presentation showed a major regression in performance that it is solved only by the use of 9.0 beta and postgis 1.5. It is in theory possible to upgrade to 1.5 of Postgis with postgres 8.3. Etienne Chove has deployed on the free server postgis 1.5 on Debian. Emilie Laffray
