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

Re: Measuring activity/problem with timeouts

--On Monday, January 17, 2005 10:33:30 AM -0800 Quanah Gibson-Mount
<quanah@stanford.edu> wrote:

> Well, if you only have issues around that time, every night, are you sure
> there isn't something happening on your network at that time?  Have you set
> up a script to do a query to "localhost" from the system, and run it every
> 2 or 3 minutes in the timeframe you are having problems to see if the that
> also experiences issues?

Actually, I forgot to mention that I do have a little job running every
minute out of cron just to run 'uptime' and append the output to the end of a
file.  During the times in question, the 1 minute load average can go up as
high as 5, but my experience with the box says that that's not sufficient to
cause timeouts.

> Also, do you collect any stats on what is happening on the system during
> that time period?  You might want to take a snapshot of "top" or some other
> indicator of system activity to see what exactly is happening on the system
> during that time period, in case the issue is with the system.

The box also houses end-user interface to our anti-spam program and there are
a number of cron jobs that run, but they do so throughout the day and there's
nothing special about the around midnight and around 4:30am time slots.  On
the other hand, from 4am until about 6am, a lot of maintenance in the form of
adding and updating entries with the heaviest demand being between 4:40am and
about 5:15am, but that doesn't correlate with the load curve.
> On a slightly separate note, OpenLDAP 2.2.17 had memory leak & cacheleak
> issues.  You might want to upgrade to the latest OL stable version
> (2.2.20).  I also have two patches to OL 2.2.20, one that fixes a problem
> with slurpd replication where changes can be lost if there is a high volume
> of changes, and one that fixes an issue that can occur with indexing using
> back-bdb.

Based on the above, I've gone ahead and upgraded.  I've also increaed the
cachesize from its default of 100k to 20000k.  I'll know tomorrow weather any
of that helped.


> "These censorship operations against schools and libraries are stronger
> than ever in the present religio-political climate. They often focus on
> fantasy and sf books, which foster that deadly enemy to bigotry and blind
> faith, the imagination." -- Ursula K. Le Guin

Rob Tanner
UNIX Services Manager
Linfield College, McMinnville OR