Re: slapd syncrepl consumer having permanent high CPU load

[John Morrissey <jwm@horde.net>]

On Fri, Mar 13, 2009 at 07:12:42PM -0400, John Morrissey wrote:
> On Thu, Mar 12, 2009 at 03:56:15PM -0700, Quanah Gibson-Mount wrote:
> > --On Thursday, March 12, 2009 1:52 PM -0400 John Morrissey <jwm@horde.net> wrote:
> > >On Tue, Mar 10, 2009 at 09:48:20AM -0700, Howard Chu wrote:
> > >>John Morrissey wrote:
> > >>> and slapd's memory footprint is increasing at ~100MB/hour. CPU
> > >>> consumption skyrockets and performance takes a nose dive. The CPU
> > >>> consumption seems to be spread across all threads; I can get another
> > >>> backtrace next time it happens, if that will help.
> > >>
> > >> Yes, please provide a backtrace.
> > >
> > > Thread 14 (Thread 0x47aeb950 (LWP 30090)):
> > > # 0  0x00007f61f7eb4007 in __db_tas_mutex_lock () from
> > > # /usr/lib/libdb-4.7.so
> > 
> > What options was BDB 4.7 built with?  And please refresh me on your OS.
> 
> It's the stock Debian packaging for BDB 4.7, on Debian lenny.
> 
> I just got around to rebuilding with --enable-posixmutexes
> --with-mutex=POSIX/pthreads; the autoconf build output differs like so:
> 
> -checking for mutexes... POSIX/pthreads/library/x86/gcc-assembly
> +checking for mutexes... (cached) POSIX/pthreads
> 
> So far (slapd's only been up for 1.5h), the DN cache is stable at ~20k
> entries, but the entry cache is at 17k entries (vs. configured 10k) and
> expanding. It seemed to plateau at ~10k for a few minutes, but then
> continued its march onward.

After ~16h uptime, slapd with this BDB had increased its DN cache to ~250k
entries after it previously appeared stable at the configured 20k entries,
and its entry cache had ballooned to ~480k entries. Its RSS was about 3.6GB
at this point, with a BDB cache size of 2GB.

john
-- 
John Morrissey          _o            /\         ----  __o
jwm@horde.net        _-< \_          /  \       ----  <  \,
www.horde.net/    __(_)/_(_)________/    \_______(_) /_(_)__