Hi! Sorry to respond so late but we've been working on other issues in last hours.
Luca Morandini-2 wrote: > > One question: how comes the load balancer switches between GWC and GS > instances ? > > Shouldn't it talk to GWC only (which, in turn, sends a request to GS in > case of a > cache miss) ? > > Could the crash be a case of "Out of permgen space" ? > > Regards, > > Luca Morandini > [http://www.lucamorandini.it] > Balancers talks to GWC and between GWC and GS, when cache misses. There are 2 GWC and 4 GS. We supposed that balancer send the same amount of request to all the instances. The balancer is configured for switching per session, ie, if a user starts to make requests to a GWC, all requests will be forwarded to it (high probability of cache hits). We configured it according to: http://docs.geoserver.org/2.0.x/en/user/production/container.html For exaple, running java with next parameters: JAVA Memory arguments: -Xms1024m -Xmx1024m -verbose:gc -XX:ParallelGCThreads=2 -Xint -XX:MaxPermSize=128m We have an OutOfMemmory Error: # A fatal error has been detected by the Java Runtime Environment: # # java.lang.OutOfMemoryError: requested 9437184 bytes for jbyte in /BUILD_AREA/jdk6_21/hotspot/src/share/vm/prims/jni.cpp. Out of swap space? # # Internal Error (allocation.inline.hpp:39), pid=3972, tid=1713052560 # Error: jbyte in /BUILD_AREA/jdk6_21/hotspot/src/share/vm/prims/jni.cpp # # JRE version: 6.0_21-b06 # Java VM: Java HotSpot(TM) Server VM (17.0-b16 interpreted mode linux-x86 ) # An error report file with more information is saved as: # /oracle/product/11.1/user_projects/domains/des1_dap/hs_err_pid3972.log ..... the pid lod: --------------- T H R E A D --------------- Current thread (0x678be800): JavaThread "[ACTIVE] ExecuteThread: '10' for queue: 'weblogic.kernel.Default (self-tuning)'" daemon [_thread_in_vm, id=4791, stack(0x66161000,0x661b2000)] Stack: [0x66161000,0x661b2000], sp=0x661afcf0, free space=13b661b2000k Native frames: (J=compiled Java code, j=interpreted, Vv=VM code, C=native code) V [libjvm.so+0x6a9262] V [libjvm.so+0x2b277f] V [libjvm.so+0x3c0b0f] C [libclib_jiio.so+0xadf31] C [libclib_jiio.so+0x8935] Java_com_sun_medialib_codec_png_Encoder_encode+0x85 j com.sun.medialib.codec.png.Encoder.encode(Ljava/io/OutputStream;Lcom/sun/medialib/codec/jiio/mediaLibImage;JI)V+0 j com.sun.medialib.codec.png.Encoder.encode(Ljava/io/OutputStream;)V+14 j com.sun.media.imageioimpl.plugins.png.CLibPNGImageWriter.write(Ljavax/imageio/metadata/IIOMetadata;Ljavax/imageio/IIOImage;Ljavax/imageio/ImageWriteParam;)V+796 j org.geotools.image.ImageWorker.writePNG(Ljava/lang/Object;Ljava/lang/String;FZZ)V+373 j org.vfny.geoserver.wms.responses.map.png.PNGMapProducer.formatImageOutputStream(Ljava/awt/image/RenderedImage;Ljava/io/OutputStream;)V+95 j org.vfny.geoserver.wms.responses.DefaultRasterMapProducer.writeTo(Ljava/io/OutputStream;)V+6 j org.vfny.geoserver.wms.responses.GetMapResponse.writeTo(Ljava/io/OutputStream;)V+5 .... Arne Kepp-2 wrote: > > During your tests, what is your cache hit ratio? > > GWC should with some fairly predictable tweaks (number of threads > supported by Tomcat, and so forth) be able to handle around 4000 tiles > per second, at which point gigabit ethernet will be saturated, for most > tiles. > > -Arne > We need to test situation with low cache hit ratios, for example, lots of locations on different areas of the map. We have made many test plans, but we are particularly interested in ratios close to 0-20% I'd rather work with Tomcat but we must work with WebLogic, which is much more inflexible. 4000 tiles per second? We don't get it in 1 minute! :S mcr wrote: > > Did you test native sql queries with a bounding box against your > oracle tables ?. > > Did you test your application servers individually ? > > Are the performance differences between the cluster members ? > > Which operating system are you using ? > Yes, queries run on DB perfectly. Servers individually crashes too. When we work with several servers,all of them have similar performance. Sometimes, only one goes down before whiile the rest are still working. We deploy in a Red Hat Enterprise 5.1 32 bits Simone Giannecchini wrote: > > A few quick questions: > > - define better what you mean for "GeoServers or GeoWebCaches instances > crashes" > - which version of GeoServer are we talking about? > - I need to know the following about your mosaic of ecw: > > -1- number of tiles in the mosaic > -2- gdalinfo on a representative tile > -3- how you configured the tile > -4- version of imageio-ext in use. > Geoserver 2.0.2 GWC and GS stop working for various reasons: - Out of Memmory - Garbage Collector - Unable to create new threads (By curiosity: we realized that many times it fell when we set a time rendering. It couldn't create the own thread which manages this timeout. Since then, we set timeout 0). I think ECW isn't a problem because without rasters layers crashes too. Anyway, we use imageio-ext-1.0.5-linux32-mrsid-ecw-lib and JAI 1.1.3. Rasters are configured as image mosaics (about 200 ecw files) -- View this message in context: http://old.nabble.com/Doubt-about-Geoserver-performance-tp29856704p29885122.html Sent from the GeoServer - User mailing list archive at Nabble.com. ------------------------------------------------------------------------------ Beautiful is writing same markup. Internet Explorer 9 supports standards for HTML5, CSS3, SVG 1.1, ECMAScript5, and DOM L2 & L3. Spend less time writing and rewriting code and more time creating great experiences on the web. Be a part of the beta today. http://p.sf.net/sfu/beautyoftheweb _______________________________________________ Geoserver-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/geoserver-users
