Re: slapd problems under stress

[Mark Valence <kurash@sassafras.com>]

The problem in this case is that too many threads are being started 
by slapd.  The test program throws search requests at the server as 
fast as possible.  Since those requests essentially happen 
simultaneously on the server, there are that many threads running (or 
trying to run) at the same time.  Most systems will bog when a 
process is trying to run many hundreds of threads, and that is what 
you are seeing.

Thread pools can fix this problem, since they can keep the number of 
simultaneous threads below some maximum.  The current build does not 
set a maximum, so behavior is similar to what's described above. 
With a new config option that I'll add, you'll be able to set this 
max threads value, which will default to something reasonable.

Mark.