Hi guys! This my first question here.

My database need a concurrency control and I am using locks.
In tests I am using 100 Threads to insert anything in database. For
each insert I do:
-lock_id
-lock_get
-txn_begin
-db.put
-txn.commit
-lock_put

This sequence for each insert is correct?
To make checkpoint or syncs I need locks and transactions too?

I am using Python (Bindings for BerkeleyDB[1]), but I believe that
operation is not different in other languages.

[1] - http://pybsddb.sourceforge.net

* Yoshio:

You don't need to use explicit locking when you use the transaction
data store. Just make sure that you set the DB_THREAD flag for the
environment and the databases when you use Berkeley DB from multiple
threads within the same process.

You might need some lock to synchronize checkpoints, log archival and
log deletion, depending on your data durability requirements.

On 13 abr, 16:29, Florian Weimer <f... (AT) deneb (DOT) enyo.de> wrote:
I will test the transactions without explicit locking.
Thank you for help.