Andrew Jorgensen andrew at jorgensenfamily.us
Tue Sep 30 17:25:59 MDT 2008

On Tue, Sep 30, 2008 at 5:09 PM, Michael Torrie <torriem at gmail.com> wrote:
> I'm not really sure.  I think there was a program called ntop, which
> gave bandwidth for each process.  But that's not the entire story.  A
> network can be nearly idle but still waiting on NFS, which causes load
> average because of the way the kernel treats NFS operations.  On the
> other hand tons and tons of connections to a service, say httpd, can
> drive up the load in a hurry.  This would likely be represented by
> saturated bandwidth, though, which ntop would show.

This is a basic LAMP box in just about every way.  No NFS.

>> Can it also be caused, for instance, by an apache thread waiting for
>> mysql to respond?  If so then how do I tell if that's the case?
> I believe that can certainly influence the load, yes.  Tuning apache and
> mysql are not things I am familiar with.  I deal mainly in file servers.

If I could be sure that was where the problem was I'd dive into
solving it.  Just gotta figure out what's really going on first.

I know they've done some improper apache tuning.  I think we're gonna
fix some of that.  I think there's still more going on though.

iostat requires a newer python than we've got.  We're looking at
vmstat but it's harder to read.

Anyone want to chime in on how to see if apache vs. mysql is where my
problem is?

More information about the PLUG mailing list