Hardware and performance

I can possibly understand the admin homepage taking long to load especially if you have thousands of courses and several categories.

But are you speaking about _All_ the admin pages? Do you have any php Caching software installed? Is your mdl_log table enormous?

Cheers,

-n

I should clarify the situation a bit. Sorry.

Moodle Cluster setup using 2 x LVS (ubuntu 9.10 64 bit) running on vmware esxi 4.0
with 3 x Windows 2003 web/real servers running apache 2.2.11 and php 5.2.8 running on VMWare Esx 2.5 HP Fiber Channel SAN setup.
all set to deliver standard theme with mod_deflate and mod_headers enabled
as well as the core mods from the recommended docs.

Moodle Data is shared on one of the web Servers as a windows share with proper access given to the web account for the apache process on the web servers

Gigabit LAN

DB is MySQL/ with myisam tables running on Windows 2003 (mysql version 5.1.34)

All servers have 2 chips (2.2 ghz dual core) and at least 2 gb of RAM except the SQL server which has 4 dual core chips and 4 gb of RAM.

Moodle 1.9.2 Build 20080711

Session handling in the database,

Internal cache for db records with record cache enabled
set to 10 and 50 as the docs say.

ADOdb output of

mysql
Parameter	Value	Description
Ratios
MyISAM cache hit ratio	99
InnoDB cache hit ratio	0	Cache ratio should be at least 90%
sql cache hit ratio	51.3
IO
data reads	1	Number of selects (Key_reads is not accurate)
data writes	0	Number of inserts/updates/deletes * coef (Key_writes is not accurate)
Data Cache
MyISAM data cache size	48M
BDB data cache size	Error:
InnoDB data cache size	500M
Memory Usage
read buffer size	65536	(per session)
sort buffer size	256K	Size of sort buffer (per session)
table cache	Error:	Number of tables to keep open
Connections
current connections	50
max connections	800

We just set the MySQL config this way late today to try to mimic a test server we had which shows almost no slowdown in performance when accessing the admin interface. The only difference was we changed our database to run on localhost on that test box, not from the DB server. (Which makes me think something is going wrong in the communication or contact with the MySQL server in my production environment)

OK. So we had APC enabled on this cluster of web servers and xcache enabled on the test server earlier today (and all this week). No change in results in either test or production environment with any PHP accelerator enabled. If anything, it seemed to slow down the page loads a bit. Which is very surprising to me.

My mdl_log table is quite large. ADOdb reads out:

mdl_log

MyISAM

10

Dynamic

753889

93

70381024

281474976710655

54381568

0

777437

Thu 31, Jul 2008, 02:19:48

Fri 20, Nov 2009, 09:49:47

Fri 20, Nov 2009, 03:11:28

utf8_general_ci

Every action is logged as far as possible

I guess I should not say it is quite large as I don't know how large it is supposed to be.

I went to this cluster to try to alleviate some apache PHP crashing issues we were having with the Quiz module, and that has been alleviated, but the Admin interface (almost every link, especially anything in the server area or users area or enrollments area) loads slowly now.

Anyone have any ideas? In my mind, this setup should rock the house, and easily fly through a DB connection and web traffic. Heck, with mod_deflate our whole template load a 40000 bytes in under 3 seconds. It load very fast, but either the PHP processing takes too long, or the database is killing us. I can't seem to find the source. BTW, the cpu usage is very low on the web server all day long, and the cpu on the DB server hardly shows any activity.

I am no expert, by any means, and I would greatly appreciate any assistance with this matter or tips for tuning.
Thanks