Problems with TileCache at zoom level 16

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Problems with TileCache at zoom level 16

Chris Holmes-2
So I'm using TileCache to pre-cache everything, and I've come to figure
out that it seems to have screwy behavior at zoom level 16.  I kept
thinking it was just me, but it turns out that at zoom 16 I get 5 by 5
areas of tiles, that have a border of 'sorry, no data available' (which
is my image for a broken link).  If you scroll around this pattern is
repeated, instead of just being some random no data available because of
load on the server or some such.

You can see this in action at:
http://sigma.openplans.org/dev/test.html?lat=40.76349&lon=-73.97426&zoom=16&layers=BT

All of manhattan at zoom level 16 is pre-cached, but there's a grid of
tiles not returning around chunks of tiles 5x5.


I'm also curious about settings to decrease how often I get tiles that
don't show up.  It doesn't appear to be a timeout (since it doesn't seem
to behave according to a set time), but when I use squid it would just
continue to wait on tiles until GeoServer loaded them.  With tilecache
if I'm at a zoom area that hasn't been pre-cached and move around
quickly then I'll often get stretches of not found tiles, instead of
just waiting to load from Geoserver.

I also sometimes even get tiles not found on areas that are pre-cached,
but it does take me moving and zooming quite a lot.

keep up the great work,

Chris



--
Chris Holmes
The Open Planning Project
http://topp.openplans.org

_______________________________________________
Users mailing list
[hidden email]
http://openlayers.org/mailman/listinfo/users

cholmes.vcf (281 bytes) Download Attachment
Reply | Threaded
Open this post in threaded view
|

Re: Problems with TileCache at zoom level 16

Christopher Schmidt-2
On Wed, Nov 29, 2006 at 05:02:42PM -0500, Chris Holmes wrote:

> So I'm using TileCache to pre-cache everything, and I've come to figure
> out that it seems to have screwy behavior at zoom level 16.  I kept
> thinking it was just me, but it turns out that at zoom 16 I get 5 by 5
> areas of tiles, that have a border of 'sorry, no data available' (which
> is my image for a broken link).  If you scroll around this pattern is
> repeated, instead of just being some random no data available because of
> load on the server or some such.
>
> You can see this in action at:
> http://sigma.openplans.org/dev/test.html?lat=40.76349&lon=-73.97426&zoom=16&layers=BT
>
> All of manhattan at zoom level 16 is pre-cached, but there's a grid of
> tiles not returning around chunks of tiles 5x5.

This is probably an issue with the threshhold in TileCache. We're going
to be working on a new release on Friday -- we'll incorporate looking
into this issue for you at that time.

> I'm also curious about settings to decrease how often I get tiles that
> don't show up.  It doesn't appear to be a timeout (since it doesn't seem
> to behave according to a set time), but when I use squid it would just
> continue to wait on tiles until GeoServer loaded them.  With tilecache
> if I'm at a zoom area that hasn't been pre-cached and move around
> quickly then I'll often get stretches of not found tiles, instead of
> just waiting to load from Geoserver.
>
> I also sometimes even get tiles not found on areas that are pre-cached,
> but it does take me moving and zooming quite a lot.

Both of these problems problems are likely with OpenLayers, rather than
TileCache. TileCache will wait forever -- the reason you're seeing 'No
data" is because OpenLayers is requesting tiles which do not fit the
tilecache boundaries. Under squid, you would still get these areas, but
you would be polluting your cache: TileCache never pollutes its cache,
exploding on you instead ;)

Patches to OpenLayers in that regard are gladly accepted, since I can't
find out where the issue is. This is part of needing better handling of
tile loading in general to ensure tiles are always on the correct grid,
and has been reported repeatedly with limited success at tracking down
the problem.

Regards,
--
Christopher Schmidt
MetaCarta
_______________________________________________
Users mailing list
[hidden email]
http://openlayers.org/mailman/listinfo/users