[Date Prev][Date Next]
Re: Progressive service degradation
On Tuesday, July 3, 2001, at 09:19 AM, firstname.lastname@example.org wrote:
I've been running some saturation tests against OpenLDAP
and have not yet been satisfied that I have things
configured correctly. The level of service is
acceptable for the first two hours, but during the third
hour, performance drops. It continues to plummet over
the course of the next 9 hours (it was an 12-hour
test). Here are the specific technical details:
Red Hat 7.1
OpenLDAP 2.0.11 - compiled from source
Berkeley db 3.1.17 - from Red Hat RPM
Pentium III 733
Pentium II 450
Here's the performance stats:
Hour Avg Time Transactions Fails Timeouts ops/sec
0100 289 123,961 37 4 34.4
0200 309 115,924 34 3 32.2
0300 486 73,837 18 2 20.5
0400 749 47,987 13 1 13.3
0500 957 37,595 20 13 10.4
0600 1,158 31,067 52 41 8.6
index objectClass eq,pres
index cn eq,sub
index sn eq
index description eq
index seeAlso eq
Only 1500 entires to cache? Lowest class on directorymark is 10,000....
I would like to know what additional configuration can
be performed in order to maintain an acceptable level of
performance over a longer period of time than two hours.
Have you verified it's not a client-side problem?
If you reload slapd when it's in a "degradation" cycle, do the
Does the server show excessive CPU load over this time?
You might want to take the approach of seeing what is actually
changing over time, to see how that's affecting the software.
Maybe a cron script to feed you performance info about the
server itself, over the course of a few hours, to see if there's
something else occuring.
email@example.com, 520-326-6109, http://www.opus1.com/ron/
The opinions expressed in this email are not necessarily those
my employers, or any of the other little voices in my head.