Patrick,

Since you are specifying DB_TXN_NOWAIT, the DB_LOCK_NOTGRANTED return
indicates that a lock is held on that key (acatually the page holding
that key). You can run db_stat -Cl on the environment to see
what locks are currently held. I don't recall any bug fixes in
this area since 4.1.24. We fixed some bugs involved with
DIRTY_READ and lock timeouts but you are not using those features.

Michael Ubell
Sleepycat Software.

ubell (AT) sleepycat (DOT) com (Michael Ubell) wrote in message news:<36b1555f.0308190849.6b451aca (AT) posting (DOT) google.com>...
Patrick earlier wrote..

One thing that occurs to me (because I hit a similar case): After you
run through the database with a cursor, are you certain that you are
closing the cursor after getting 'NOTFOUND' (i.e. after reading the
last record)? It's non-intuitive but the cursor does still have a
lock on the last page even if you interpret "NOTFOUND" to mean 'have
run off the end of the database'. (I believe the lock is held in
order to support Dbc::get(DB_PREV) on a cursor that has returned
NOTFOUND to a get(DB_NEXT).)

Hope this helps.

regards,

Chris