JRE/JVM crash in centera sdk on websphere

Hi,

We have a Centera application running on websphere on solaris that basically pulls images out of centera and display them for users, it works reasonably well, but once every week or so, it is crashing the websphere application from the native centera code.  The hs_err_log file contains:

Current thread (0x040a4c00):  JavaThread "Blob Reader" daemon [_thread_in_native, id=80943, stack(0xa5080000,0xa5100000)]

siginfo:si_signo=SIGSEGV: si_errno=0, si_code=1 (SEGV_MAPERR), si_addr=0x00000035

....

Stack: [0xa5080000,0xa5100000],  sp=0xa50ff060,  free space=508k

Native frames: (J=compiled Java code, j=interpreted, Vv=VM code, C=native code)

C  [libc.so.1+0xc3d68]  mutex_lock_impl+0x68

C  [libFPLibrary32.so.3.2.705+0x68a94]  fp_mutex_lock+0xc

C  [libFPUtils32.so.3.2.705+0x359f8]  void FPMutex::lock()+0xc

C  [libFPCore32.so.3.2.705+0x94e04]  void FPTag::BlobRead(FPBasicGenericStream&,long long,long long,int)+0x890

C  [libFPCore32.so.3.2.705+0x72698]  void _FPTag_BlobReadPartial(FPTag*,FPBasicGenericStream*,const long long,const long long,const long long)+0x148

C  [libFPLibrary32.so.3.2.705+0x55410]  void FPIntercept_Router::FPTag_BlobReadPartial(FPTag*,FPBasicGenericStream*,const long long,const long long,const long long)+0x58

C  [libFPLibrary32.so.3.2.705+0x45b40]  FPTag_BlobReadPartial+0xcc

C  [libFPLibrary32.so.3.2.705+0x5bb0c]  Java_com_filepool_natives_FPLibraryNative_FPTag_1BlobReadPartial+0x8c

j  com.filepool.natives.FPLibraryNative.FPTag_BlobReadPartial(JLjava/io/OutputStream;JJJ)V+1318301324

j  com.filepool.natives.FPLibraryNative.FPTag_BlobReadPartial(JLjava/io/OutputStream;JJJ)V+0

j  com.filepool.fplibrary.FPTag.BlobReadPartial(Ljava/io/OutputStream;JJJ)V+96

j  com.filepool.fplibrary.FPTag.BlobRead(Ljava/io/OutputStream;J)V+7

j  com.filepool.fplibrary.FPTag.BlobRead(Ljava/io/OutputStream;)V+3    

So, it appears that the crash originates inside the centera sdk.

I have not been able to duplicate these by pulling up the same files, so I don't believe its something special about the files its working on.  I don't think it is strictly load related, and we are not close to exceeding the 100 maxconnection limit when this occurs so I'd guess its a timing or synchronization (possibly combined with load) type of issue.

Notes about what we're doing.

We're using the default centera sdk settings  except for collisionavoidance=1 and openstrategy=0.

We retrieve a clip with up to 1000 images in it (usually more like 50),  and then retrieve the specific image by looping through the tags until the name matches (looping through all the tags to check for a match seems inefficient, but is the std approach as far as I can tell).

We read the blob directly from centera into a piped stream to the web client, no file caching is done.

Any suggestions are appreciated?

Thanks,

Roger