hello all,

in the "Berkeley DB transaction processing with c++" reference it is
said that:
"A transaction holds any locks that it obtains until the transaction is
commited or aborted"
(Lock Lifetime, pg.31)

it is also said that:
"In Addition, for cursors you can use read uncommitted isolation, which
causes the cursor
to release its read locks as soon as it is done reading the record (as
opposed to holding
its read locks until the cursor is closed" (Avoiding Blocks, pg. 33)

what about transactional cursors?
when i use a transactional cursor with READ COMMITTED isolation, is the
lock held until the transaction is committed or until the cursor has
finished its current read operation (moved to the next record)?

in addition, i find it pretty hard to debug my application in order to
see how deadlocks occur.
is there any utility that I can use to see all the locks that are held
on the database?

thank you very much.
moshe levi.

Moshe,

DB_READCOMMITTED locking releases locks as soon as possible. That is
as soon as the cursor leaves the page.

You can use the -C flag to db_stat to see what locks are held. Please
see:
http://www.sleepycat.com/docs/utility/db_stat.html

Michael Ubell
Sleepycat Software