[Date Prev][Date Next]
Re: 2.1.17 performance issues
On Tue, 15 Apr 2003, John Madden wrote:
> >> You were probably running LDBM before with 2.0.x. You may want to try
> >> LDBM again if you don't need the transactional integrity of BDB...
> > We use BDB with a particularly large DB (close to 300,000 people with
> > 20+ attributes each), and have no performance problems to speak of.
> > Our systems are heavily utilized 24/7. There are some basic
> > configuration pieces to BDB that can help improve performance. I would
> > be curious to know if your slowdowns happen to coincide with write
> > activity. Also, have you looked at your BDB stats? Perhaps you have
> > not allocated a large enough cache.
> I'll take a look at these things in the future. For now, I've rolled back
> to 2.0.27 on Linux. Just after I sent the original post, things got so
> bad that slapd eventually didn't even respond to a -INT, and I eventually
> ended up with corruption (slapcat won't even run to completion).
Is your slapd taking ages to complete queries? I'm seeing the same
behaviour after a certain number of queries. I'm still investigating it,
If isolated the cause of the problem. It's has something to do with
back-bdb. It might be bdb itself since is has a 'simple' locking method.