[OSM-dev] My Maps Appear to be Predicting Global Warming.

Richard Ive richard at xanox.net
Mon Feb 22 10:03:36 GMT 2010


Hi,

It's just a subset of tiles.

The os is 32bit version of Centos5 - It is a virtual server running on an 8
core / 32gb ram machine.

I ran render_old with the default flag, however this doesn't seem to have
had any affect.


On 19 February 2010 18:59, Jon Burgess <jburgess777 at googlemail.com> wrote:

> On Fri, 2010-02-19 at 14:40 +0000, Richard Ive wrote:
> >
> > On 19 February 2010 14:24, John Smith <deltafoxtrot256 at gmail.com>
> > wrote:
> >         On 19 February 2010 22:01, Lennard <ldp at xs4all.nl> wrote:
> >         > The one major cause of flooded tiles is when mapnik doesn't
> >         process the
> >         > coastline shapefiles. At z10, it switches from the lowzoom
> >         shoreline_300
> >         > to the full resolution processed_p shapefile. So focus your
> >         effort on
> >         > that one.
> >
> >
> >         Also these shape files are only updated intermitently, so even
> >         if you
> >         make a change to fix problems with coastlines it might take
> >         weeks or a
> >         month before the changes will be reflected in OSM tiles.
> >
> > Sorry guys, not sure I'm following here.
> >
> > I understand what you mean by shape files, but I haven't changed
> > anything on the server at all in the 3 months I've had it up and
> > running. The only thing that has changed is the database, because I
> > import daily change sets into the gis database and even that is on a
> > completely different physical machine. Why should my tiles be
> > affected, but osm.org's not?
> >
> > Cheers for the replies so far! Hopefully I'll catch on soon.
>
> Do all the high zoom tiles have the same issue or just a subset?
> Are you running on a 32 bit machine or using a 32 bit renderd process?
>
> Since the shapefiles are quite large it is possible that you run into
> problems with running multiple rendering threads in parallel. Mapnik
> tries to mmap() the file in each rendering thread and will quickly run
> out of space in a 32 bit process. Try running with a lower "-n" value,
> perhaps try just 1 to start with, then only one thread will be actively
> rendering at a time.
>
> Most modern multi-core systems will run a 64 bit OS and do not have this
> issue.
>
>        Jon
>
>
>


-- 
Regards,
Richard Ive.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/dev/attachments/20100222/954ffcd7/attachment.html>


More information about the dev mailing list