[Page 1, 2 Next]

Topic: Need some help advice [Stress testing & Performance] Apa

[1]

Topic: Need some help advice [Stress testing & Performance] Apa

[dke]

Hi all,

My setup is as follows...

Software:

Windows XP SP2 Fully updated, fully optimized with all useless services turned off.
Apache: 2.0.59
mySQL: 5.0.45
PHP: 5.2.3

Hardware:

CPU 3.2ghz P4 (Northwood)
RAM: 2GB PC3200 DDR
HDD: Seagate Barracuda 7200rpm SATA2

Issue:

Im running a image gallery on this server, and when i get around 2-5 users browsing images apache starts running sluggish.

I've tried to use newer versions of apache without any luck (actually it got worse)

I'm a newbie when it comes to caching and i think thats where my solution might be lie, as i've heard that apache is bad on hosting images raw without any caching.

On PHP side ive installed eAcellerator and it does its job well (loadtime reduced)

What i need some advice with is how i could do performance tests using the "AB" binary which is included in the apache package. Ive done some tests and the results seem frightening, but i have nothing to compare with so i don't know if they are bad or just average.

This is what i tested:

ab -kc 10 -t 30 http://localhost/

[Steffen]

Just done the test on apachelounge.com. The page at localhost is a index.html.

Quite different figures then you, I noticed I have Keep-Alive requests: 30713 and you 0


Software;

Windows XP SP2 Fully updated, fully optimized with all useless services turned off.
Apache: 2.2.4 (from Apachelounge)
mySQL: 5.0.46
PHP: 5.2.3

On the same box running Mail server, Sambar server, BIND and see www.apachelounge.com/siteinfo.html


Hardware:

CPU 2.8ghz P4 (Northwood)
RAM: 2GB PC3200 DDR
HDD: 2X Seagate ST340014A (40 GB, 7200 RPM, Ultra-ATA/100)


ab -kc 10 -t 30 http://localhost/

[dke]

Hi Steffen,

That's impressive statistics compared to mine, and you're running tons of stuff on it also! Maybe i've finally located my issue.

Would you be so kind to share some of your settings in httpd.conf?

Timeout
Keepalive (on obviously?)
MaxKeepAliveRequests
KeepAliveTimeout
ThreadsPerChild
MaxRequestPerChild

Also any other setting you feel might be nessesary?

I'll try the apachelounge apache2.2.4 build tomorrow again, to see if i maybe did something wrong with it the first time i tested (i felt that i got degraded performance with it)

Thanks for the quick reply!

[dke]

Hi again,

Ive tested some things now and the issue lies with PHP somewhere.

When PHP is disabled i get these statistics:

[Steffen]

Running a virus scanner which is scanning .php files ?
Not used php modules commented out ?
When you run as fastCGI is the httpd.exe also consuming 70% CPU or it is then going to php_cgi ?

I have:

Keepalive on
MaxKeepAliveRequests 100
KeepAliveTimeout 20
ThreadsPerChild 250
MaxRequestPerChild 0

25 requests per second is not bad at all. You can run with this a very busy site. compare it with www.apachelounge.com/stats/apache-taccesses-week.png , it's far lower then 25/sec, so I have plenty of room to grow.

Steffen

[dke]

[dke]

ive configured steffens build of 2.2.4 using fcgi now and ive noticed a heavy improvement when changeing some of the fcgi settings.. now i actually get some load on my cpu, mysql is up to 40-60% cpu usage and page is running lots smoother, can handle alot more now..

Im pretty happy with as it is now, but id though id ask you experts if theres anything i could improve in mysql to make it run even better?

below is taken from info script that steffen made

[Steffen]

I see you set the Query cache in MySQL. It's not always helping, what happens when you switch it off.

Steffen

[dke]

To turn of query cache had a drastic effect on the benchmark. From 28 requests/second to 2 requests/second. It's clearly mySQL which is slowing apache down, but i don't know how to give mySQL more juice. I have given it tons of memory, around 1,2gb, maybe i'm not allocating it wisely?

I have the majority of memory dedicated to key_buffer_size.

[tdonovan]

I noticed that you used 20,048 Connections in 21,280 seconds of Uptime. That's about one new connection per second, which seems like a lot.

A few things you might check:

* If Coppermine Gallery uses php_mysql, check your PHP.INI [MySQL] section and make sure that it has mysql.allow_persistent = On.
If it uses php_mysqli, then persistent connections should always be used AFAIK.

* Check in httpd.conf that your fcgid processes live long enough. The default for ProcessLifeTime is 3600 seconds (an hour).
Make sure this isn't set to some small value.

* If Coppermine Gallery uses php_mysql (not php_mysqli) and MySQL is on the same machine, try using named pipes instead of port 3306 for MySQL.
Start MySQL with the --enable-named-pipe switch and make sure that mysql.default_host = localhost in PHP.INI.
This is a long shot, but who knows - maybe your port 3306 connection is slow for some reason.

-tom-

[dke]

[dke]

For some strange reason i cannot get named pipes to work. I've read some FAQ on mySQL and it says named piping only works with windows 2000 and higher.. im using XP so it should work (??)

Anyhow, i've configured this way:

In MY.INI

[Steffen]

Once I had it working. When I recall you have to use in Wondows "." as host in your php application.

Form the manual:

On Windows, the hostname . causes the client to connect to the local server using a named pipe



Steffen

[tdonovan]

It's only default_host you need to change for named-pipe in PHP.INI, not default_socket in PHP.INI or socket in MY.INI.

Try executing the MySQL server from the command-line with --enable-named-pipe --console to see if it is capable of named-pipe.
The MySQL Enterprise version is usually built without named-pipe support, but the Community version usually has it.
If it says unknown option '--enable-named-pipe' your server doesn't have it.

named-pipe works OK for me on w2k with Apache-2.2.4/PHP-5.2.2/MySQL 5.0.46 and default_socket = localhost in my PHP.INI.
If you get it working, request mysql_info.php and look for named_pipe = ON.

I wouldn't struggle with named-pipe too much - I find it makes a very small difference.

More important, have you figured out why you are creating so many connections?
Since you only used a max of 11 simultaneously (Max_used_connections), it doesn't seem reasonable that you created 20,000+ new connections in 6 hours of runtime.

-tom-

EDIT: Maybe Coppermine Gallery uses php_mysqli instead of php_mysql.
php_mysqli doesn't use persistent connections, so that would explain why you create so many new MySQL connections.
If this is the case, I expect there isn't much you can do about the number of connections - it's probably best to move on to tuning something else in MySQL.
-t-

[dke]

[tdonovan]

re: [mysqld-nt] - if you mean the sections in my.ini - I don't think there is a [mysqld-nt] section, just a [mysqld] section, regardless of which executable you run.

The MySQL startup display always shows socket '' on Windows. It will not show the name of the pipe.

The host name when you want to use named pipes on Windows is "." (a single dot) per the MySQL client API doc.

To use named pipes you need to:

1. set enable-named-pipe (and, if you want, skip-networking) in my.ini in the [mysqld] section.

2. set socket to the same name in my.ini, in both the [client] and [mysqld] sections.

3. set mysql.default_socket to the socket name in php.ini

4. change the host name from localhost to . (a single dot) in Coppermine.

This may not buy you much performance. There are varying reports about whether named pipes are faster or slower than TCP on Windows.

You could try shared-memory instead of enable-named-pipe. This should not require any config changes to PHP or Coppermine to work if Coppermine uses host name "localhost".

-tom-

[dke]

Thank you for the help, however i still cannot configure this named pipe.

Documentation seems really bad for windows users.

I've done the exact thing you say, and when i do it i get a message from "CreateNamedPipe":

[tdonovan]

I think you misunderstood.

The host name (somewhere in Coppermine configuration) should be dot.
This change, from localhost to . (dot), won't be in either php.ini or my.ini - it will be in some Coppermine configuration file.

The socket name cannot be dot. Try something like "MySQLpipe" (without the quotes).
The socket name must be the same in all three places: two sections of my.ini and once in php.ini.
It is case sensitive, so make sure it is exactly the same in all three places.

-tom-

[dke]

Thank you, its working properly now!

It was a slight improvement, but i still get really high respons times out of apache when getting around 30req/sec

I'm going to try windows 2003 server r2 just to see if there is a operating system issue, thanks for all your help!

EDIT:

Tested a server operative system, and i didn't get any increase or decrease in performance, so i guess its really a coppermine issue with the queries it sends the database. Thanks for the help fixing the pipe though, it really got better, about 5 more requests/sec!

[tdonovan]

Your posts made me curious, so I downloaded Coppermine Gallery version 1.4.12 and gave it a try with Apache 2.2.4 (from Apache Lounge), PHP 5.2.2 via mod_fcgid, and MySQL 5.0.46 on the same machine.

My Windows 2000 machine (3.4ghz 2gb pentium) is roughly similar to yours.

Using XDebug shows that most of the time is spent executing PHP code, so an optimizer like eAcelerator is definitely warranted. I used XCache which is similar to eAccelerator, and I got a similar performance increase. It increased my throughput from 10 to 30 pages/sec using Apache Bench (ab.exe) against /coppermine/index.php.

It appears that ab is not a very fair measurement tool for an application like Coppermine, where users select images from thumbnails and can vote on various images. There are two reasons:

1. Using ab causes every request to start a new Coppermine user session because ab doesn't retain cookies.

2. No real users would make a thumbnail selection or choose a picture rating in milliseconds.

Instead, I recorded a script using the old, free (but still very useful) Web Application Stress Tool. I recorded a script (incluing browser pauses) as I wandered through the app as fast as I could - selecting thumbnails in about 1/2 second and voting immediatialy on pictures. Faster than any real user would go - but still more realistic than hammering the app with ab. I also configured WAST to add a random 5-50ms pause to each request, so the test would simulate users going at slightly different speeds.

I found that I could support 100 such fast-acting users comfortably, with all response times below 1 sec. CPU use would occasionally get near 100%, but the whole machine (and Coppermine) stayed very responsive throughout the test.

FYI - here are two other tuning things I tried. They helped a little - but not really enough to bother with unless you want to squeeze every last bit of performance out of Coppermine:

* Changed the MySQL storage engine for the session table (cpg1411_sessions) to an in-memory table and added an index to speed up the search for expired sessions:

Quote:

mysql> DELETE FROM cpg1411_sessions; mysql> ALTER TABLE cpg1411_sessions ENGINE MEMORY; mysql> ALTER TABLE cpg1411_sessions ADD INDEX (time);

The disadvantage to this is that Coppermine user sessions won't be retained over a MySQL shutdown and restart.

* Altered coppermine\include\functions.inc.php line 181 to call mysql_pconnect instead of mysql_connect. This makes Coppermine re-use persistent connections instead of creating a MySQL connection for each request. Probably helpful if your MySQL is on another machine, but it didn't do a lot for me. It did cut way down on the Connections shown in mysql_info.php.

Hope this stuff is useful (or if not useful - at least interesting) to you.

It's still a guess how many "real life" users I could support - but I would guess 200-300 unless all my users had the curious ability to make photographic decisions in fractions of a second

-tom-