[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: (ITS#6200) slapd crashes under load w/ syncrepl

On Mon, 24 Aug 2009, Aaron Richton wrote:

> What does this look with top and/or in dmesg over the run time? Is it a 
> simple out-of-memory? Definitely a bit of a gross method, but what's ls -lh 
> core show for size?

I haven't been able to monitor it constantly over its lifespan, as it 
takes anywhere from 24 to 36 hours of our load before failure, but what I 
have observed is that slapd continually grows (slowly) over the course of 
the test until it can't malloc any more.  Core sizes run around 640Mb. 
The largest '__db' element is 1.9Gb.

So, slapd behaves as if there's a slow memory leak, somewhere.  Unless the 
delta syncrepl mechanism (on producer) has an insatiable memory appetite.
Our test suite is heavy on updates, including mods to large-memeber groups
(with 120K+ members).

The consumer slapd never dies, and more or less keeps up with the change 
load from the producer.

> (If so, is it warranted given your load or is there a leak, etc etc...and of 
> course make sure you're up to date on the surrounding packages, OpenLDAP 
> isn't the only thing that can leak.)

db-4.7.25 (patch level 4)
openssl 0.9.8e
Red Hat Enterprise Linux Server release 5.4