Hi, oh, shame on me. Of course I meant "FcgidInitialEnv QGIS_SERVER_PARALLEL_RENDERING True"! Please excuse me and thank you for the hint with the log level. I will try it.
Burghardt > -----Ursprüngliche Nachricht----- > Von: David Marteau [mailto:dmart...@3liz.com] > Gesendet: Montag, 9. Oktober 2017 12:19 > An: Scholle, Burghardt (15-3 GIS) > Cc: qgis-developer@lists.osgeo.org > Betreff: Re: [QGIS-Developer] QGIS Server 3 status > > Hi > > According to the source > https://github.com/qgis/QGIS/blob/master/src/server/qgsserversettings.cp > p#L47: setting QGIS_SERVER_PARALLEL_RENDERING to False or not setting > QGIS_SERVER_PARALLEL_RENDERING is the same thing as the default value > is 'false'. > > Using the appropriate log level you can get metric on the rendering time > from qgis: imho, comparing this metric would be a good starting point putting > aside external configuration (except for the postgres connection) and isolate > any bottleneck. > > David. > > > Le 9 oct. 2017 à 11:53, <burghardt.scho...@stadt.wolfsburg.de> > <burghardt.scho...@stadt.wolfsburg.de> a écrit : > > > > Hi David, > > > > I will answer you directly and give you the desired information. I've done > some tests with a sample application, which you can see under the URL > https://lxgis1.stadt.wolfsburg.de/qgis_wcl/index.html. > > The map topic "Stadtplan" is based on the qgis server 2.14.19, the > "Stadtplan (Dev)"-topic is based on qgis server 2.99 (codeversion 825a323). > Subsequent tests I've done with the layers "Luftbilder" and "Infrastruktur". > The former are geotiff's, the latter are generated from a PostGIS database. > As client I use a customized version of the "OL3 Mobile Client" from > Sourcepole. > > > > If I look at the access times in the developer tools of the Firefox, they > speak a clear language: > > - Topic "Stadtplan" -> moving the map, zoom in / zoom out takes about > > 0.5 s on the intranet (!) > > - Topic "Stadtplan (Dev)" -> moving the map, zoom in / zoom out takes > > about 1.2 s with the setting "FcgidInitialEnv > > QGIS_SERVER_PARALLEL_RENDERING False" in my apache config > > - Without setting "FcgidInitialEnv QGIS_SERVER_PARALLEL_RENDERING > > False" the access takes about 0.7 s > > Result: without using multiprocessing / multithreading, the QGIS server is > faster under these test conditions, but still clearly slower than the older > version! > > > > Hope this description helps > > > > Burghardt > > > > *************** > > > > Burghardt Scholle > > mailto:burghardt.scho...@stadt.wolfsburg.de > > > >> Hi, > >> > >> Do we have metrics on this ? Does it mean that > QGIS_SERVER_PARALLEL_RENDERING=False is also slower than previous > version ? > >> > >> > >> David M. > >> > >> > >> > >>> Le 4 oct. 2017 à 14:15, Matthias Kuhn <matthias at opengis.ch> a écrit : > >>> > >>> Hi, > >>> > >>> I had some questions from people interested in performance. > >>> > >>> I advised them to set the following > >>> > >>> * QGIS_SERVER_PARALLEL_RENDERING=True > >>> * Trust project flag > >>> > >>> Apparently parallelized rendering works (system stats) but it was > >>> reported to be slower than older versions. > >>> > >>> Is there something else that needs to be done? Is there more > >>> information I can ask for / give? > >>> > >>> Thanks a lot > >>> Matthias > >> > > > > > > _______________________________________________ > > QGIS-Developer mailing list > > QGIS-Developer@lists.osgeo.org > > List info: https://lists.osgeo.org/mailman/listinfo/qgis-developer > > Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-developer _______________________________________________ QGIS-Developer mailing list QGIS-Developer@lists.osgeo.org List info: https://lists.osgeo.org/mailman/listinfo/qgis-developer Unsubscribe: https://lists.osgeo.org/mailman/listinfo/qgis-developer