Geoserver exceptions

[Bertrand, Matt]

I've been noticing some common and not-so-common exceptions occuring in WorldMap's geoserver instance, and was wondering if maybe one of you could shed some light on what the causes might be.  This might be more suited to the geoserver list but I thought I'd try here first.

The most serious, which has occurred today and yesterday, happens when many of the layers in GeoServer suddenly become inaccessible.  Each time this has happened has been immediately a layer upload (but there have been many uploads where everything went fine).  The layers show up as tiles of 'NullPointerException' on the map.  Geoserver has to be restarted.  The uploaded layer which may have triggered the problem is fine after the restart, and there's no errors indicated in the django log (http://pastebin.com/et809HQf).  The stack trace is here: http://pastebin.com/sgb2KLWu

Occasionally when a layer is first rendered, one of its tiles displays an exception: "org.geoserver.ows.ServletDebugException null"

Stacktrace: http://pastebin.com/HKRyLk4H

I also noticed occasional large batches of the following exception in the log but I'm not sure what the consequence was (if anything) to the enduser:

java.lang.RuntimeException: javax.imageio.IIOException: Couldn't seek!

http://pastebin.com/Z6gi6ZDk

Finally, there's a common exception which has no obvious detrimental effect but occurs on almost every shapefile upload:

15 Jul 04:16:05 WARN [shapefile.shp] - Could not open the .shx file, continuing assuming the .shp file is not sparse

java.io.IOException: Wrong version, expected 9994, got 16777216

Sometimes the '16777216' is '-2130706432' instead.

http://pastebin.com/hwNAkyDb

[David Winslow]

Some of these issues I haven't seen before.  I put some comments inline where I had comments to make.

On Tue, Jul 19, 2011 at 2:07 PM, Bertrand, Matt <mber...@cga.harvard.edu> wrote:

I've been noticing some common and not-so-common exceptions occuring in WorldMap's geoserver instance, and was wondering if maybe one of you could shed some light on what the causes might be.  This might be more suited to the geoserver list but I thought I'd try here first.

The most serious, which has occurred today and yesterday, happens when many of the layers in GeoServer suddenly become inaccessible.  Each time this has happened has been immediately a layer upload (but there have been many uploads where everything went fine).  The layers show up as tiles of 'NullPointerException' on the map.  Geoserver has to be restarted.  The uploaded layer which may have triggered the problem is fine after the restart, and there's no errors indicated in the django log (http://pastebin.com/et809HQf).  The stack trace is here: http://pastebin.com/sgb2KLWu

I think this is a bug in GeoServer, but as you mention it does not survive a restart and is pretty tricky (in my experience) to reproduce. I'm not aware of a workaround to avoid it (and it will be hard to test a workaround anyway since it's tough to reproduce the issue in the first place.) 

Occasionally when a layer is first rendered, one of its tiles displays an exception: "org.geoserver.ows.ServletDebugException null"

Stacktrace: http://pastebin.com/HKRyLk4H

I also noticed occasional large batches of the following exception in the log but I'm not sure what the consequence was (if anything) to the enduser:

java.lang.RuntimeException: javax.imageio.IIOException: Couldn't seek!

http://pastebin.com/Z6gi6ZDk

Finally, there's a common exception which has no obvious detrimental effect but occurs on almost every shapefile upload:

15 Jul 04:16:05 WARN [shapefile.shp] - Could not open the .shx file, continuing assuming the .shp file is not sparse

java.io.IOException: Wrong version, expected 9994, got 16777216

Sometimes the '16777216' is '-2130706432' instead.

http://pastebin.com/hwNAkyDb

We can change the default configuration for shapefile layers created via GeoNode to have GeoServer regenerate .shx files if it cannot read the existing ones.  The .shx file is just an index so it can be fully regenerated from other Shapefile components.

Hope this helps.

--
David Winslow
OpenGeo - http://opengeo.org/

[Bertrand, Matt]

I think this is a bug in GeoServer, but as you mention it does not survive a restart and is pretty tricky (in my experience) to reproduce. I'm not aware of a workaround to avoid it (and it will be hard to test a workaround anyway since it's tough to reproduce the issue in the first place.) 

Thanks David, have you seen this happen with GeoNode as well?  Or even a regular GeoServer instance?  That would help me rule out any changes I made for WorldMap.

[David Winslow]

I have only seen this issue in GeoNode sites but I am not sure we should rule out it being a GeoServer bug.

--
David Winslow
OpenGeo - http://opengeo.org/