Hi,

1) I'm using BerkeleyDB 4.4.20 through Java. That is, I access the DB
through the Java Swig wrappers.

2) There is only one thread accessing the DB in my application.

3) My DB is created as follows:

dbConfig = new DatabaseConfig();
dbConfig.setAllowCreate(true);
dbConfig.setSortedDuplicates(true);
dbConfig.setType(DatabaseType.BTREE);
incidence_db = env.openDatabase(null, INCIDENCE_DB_NAME,
null, dbConfig);

4) A fair amount of data gets added to this 'incidence_db' where each
key may potentially have many duplicates.

5) Both keys and values are 16 byte UUIDs.

6) I tried inserting the key/value on which it hang using a reduced
data set and it worked. So it doesn't seem dependent on the actual
key/value pair.

7) Same problem with Berkeley DB 4.3....

Any indication how I may track down the issue?

Thanks a lot in advance!
Borislav

Hi Borislav,

Is the environment configured for locking? If so, can you check
whether the thread blocks waiting on a lock (using 'db_stat -Co' in the
environment directory).

Alternatively, can you attach a debugger to the process when it hangs
and get a stack trace?

Regards,
Michael.

Hi Michael,

Thanks! Seems like there lies the problem:

Locks grouped by object:
Locker Mode Count Status ----------------- Object
---------------
33 READ 1 HELD incidencedb page
58
4 READ 1 HELD incidencedb page
58
2b WRITE 1 WAIT incidencedb page
58

2b READ 1 HELD incidencedb page
1

it's waiting on a locker, but why would that be? The environment is
configured for locking because it will eventually be accessed from
multiple threads (even though in this particular instance, it isn't).
Here is the configuration:

EnvironmentConfig envConfig = new EnvironmentConfig();
envConfig.setAllowCreate(true);
envConfig.setInitializeCache(true);
envConfig.setInitializeLocking(true);
envConfig.setInitializeLogging(true);


Could then this mean that I haven't closed some cursor on that DB
properly?

Thanks!
Borislav