[WBEL-users] Hello, and possible LAN/NIC issue?

Kirby Bohling kbohling@birddog.com
Tue, 6 Jul 2004 12:35:16 -0500


On Tue, Jul 06, 2004 at 11:20:13AM +0100, Nigel Kendrick wrote:
> Greetings - first poster here so be gentle on me!
> 
> Using WBEL with great success but come up against a small issue with
> performance that's driving me mad...
> 
> I am running eGroupWare on a PIII-750 with 256MB RAM. The Web pages seem
> to be served very slowly - for example, the PHP page generation
> indicator on eGroupware says 'page generated in 0.nnn seconds', but the
> Web pages take sometimes 7-10 seconds to turn up on the client PC -
> we're connected directly on the same switch.

Okay, there are several ways to attack this:

1.  Start using timing wget on the local host on the slow machine:

/usr/bin/time wget http://localhost/the/url/here

2.  Start using timing wget on the local host on the fast machine:
/usr/bin/time wget http://localhost/the/url/here

Now, if you see a large difference, it's pretty much a configuration
issue on the slow machine.  If you don't, something else is going
on.

Try running timing from the same client both httpd servers:

[client] $ /usr/bin/time wget http://fastmachine/the/url/here
[client] $ /usr/bin/time wget http://slowmachine/the/url/here

Now, if they are both slow, it could be logging is slow.  It could
be that that the reverse lookup is slow on one machine, but not the
other (or one of them could be setup not to do reverse lookups).
You can watch the logs to see if when you click start the
transaction how long it takes for Apache to log the result.

You can use tethereal to check things out, to see if it's if it's
just the webserver acting fast or slow.  I'd watch run tethereal on
both the client and server end to see what both of them are doing.

It could be any number of things.  You can do simple things like
ping the fast and slow machines to see if there is any difference
there.  I'd ping in both directions.  You check check the errors on
the ethernet interface they are using. 

We've had problems like this, and it turns out that it was actually
an LDAP authentication issue.  If you shut off all authentication it
ran lickty split.  If you turned the authentication on, it took
forever.  Turned out it was the LDAP server doing the username
lookup on an un-indexed key.

	Thanks,
		Kirby