[SAC] ProjectsVM Upgrade Problem

[Alex Mandel]

On 02/06/2012 01:44 AM, Markus Neteler wrote:
> On Fri, Feb 3, 2012 at 10:59 PM, Markus Neteler <neteler at osgeo.org> wrote:
>> While it seems that all web servers on the projectsVM are back,
>> the apache performance (while the CPUs are lazy...) was very low.
>>
>> I have tweaked the number of workers etc in apache2.conf, now
>> it seems to be fast again.
> 
> ... only temporarily :-(
> 
> I have taken liberty to stop srorgapache, now the performance is sort of back.
> 
> Perhaps srorgapache should go to a different machine since it is slowing
> down foss4g, gdal, geos, geotools, grass, grass wiki,  mapbender,  mapserver,
> openlayers,  remotesensing.org, and  tilecache.
> ?
> 
> Markus
> _______________________________________________
> Sac mailing list
> Sac at lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/sac

It's quite a popular site so we can't leave it off indefinitely (It
appears to be on again). I did manage to figure out how to run the site
via gunicorn instead of apache but in testing it's seems the proxy by
the main apache is what's causing the most slow down and possibly load
on the main apache. I'm not sure the load on the cpu/ram is the culprit,
but I do have an idea to test it.

In the meantime I'd like to run an experiment for a week running sr.org
on gunicorn instead of the srorg apache to see if things behave any
better? Any objections, concerns or questions?

So maybe moving to another server is the answer, but we'd need to
identify what server would be good for that.

The long term interesting idea with running wsgi stuff behind gunicorn
or uWSGI is the potential to swap out apache for nginx (when needed) and
the possibility of using supervisord to keep servers prone to crashing
from requiring human intervention to restart web services.

Thanks,
Alex