RT 3.0.10 performance

RT Users
1

I feel your pain.

I have a theory…

Care to send the output of:

cat /proc/sys/kernel/shmall
cat /proc/sys/kernel/shmmax
/usr/bin/ipcs -a
/usr/bin/uname -a

to either the list or to me directly?

Arnold Cavazos, Jr. abcjr at abcjr . netOn Tue, Apr 13, 2004 at 06:12:02PM -0700, Dan O’Neill wrote:

RT users,

I’d like to chime in on the RT performance issue. First, understand that
I think RT is the best ticketing system that I’ve ever had the pleasure of
using. Second, there is something about RT that is slow and it’s not the
database. Third, this is a long message so I apologize in advance, but I
want to present all the necessary information.

And last, I’m looking to work with someone and try different things to
make this product faster. I’m willing to do some profiling work as well.

Configuration information:
Hardware:
Dual 1GHZ Pentium pro
1GB RAM
1000rpm 36GB scsi
100mbps Intel NIC

Operating system:
Fedora Core 1
All RT required modules installed
EXT 3 file system on all partitions

Output from top:

17:53:20 up 8 days, 9:18, 1 user, load average: 0.00, 0.00, 0.00
72 processes: 71 sleeping, 1 running, 0 zombie, 0 stopped
CPU states: cpu user nice system irq softirq iowait idle
total 0.0% 0.0% 2.8% 0.0% 0.0% 0.0% 197.0%
cpu00 0.0% 0.0% 2.9% 0.0% 0.0% 0.0% 97.0%
cpu01 0.0% 0.0% 0.0% 0.0% 0.0% 0.0% 100.0%
Mem: 1032040k av, 707972k used, 324068k free, 0k shrd, 103720k
buff 230444k active, 378380k inactive
Swap: 514072k av, 0k used, 514072k free 211692k cached

Apache process size:

PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME CPU COMMAND
11092 root 16 0 23176 22M 10796 S 0.0 2.2 100:09 1 httpd

Users:
Maybe 5 people use it on a daily basis

of Tickets:

 150 as of 3/4/04

Now, look at a queue with 20 NEW tickets. Select the NEW tickets from
the queue list, press next, and Apache starts to use a tremendous amount
of CPU to present the ticket. It’s not just this sequence either, it’s any
sequence of events that result in information being retrieved.

PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME CPU COMMAND
11092 root 16 0 23176 22M 10796 S 96.4 2.2 101:19 0 httpd

This is a regular occurance. The system is dedicated to RT. Here is the
relevant Apache configuration information:

Apache/1.3.29 (Unix) mod_perl/1.29 mod_ssl/2.8.16 OpenSSL/0.9.7a

MinSpareServers 5
MaxSpareServers 10
StartServers 5
MaxClients 150
MaxRequestsPerChild 0

<VirtualHost *:80>
ServerName rt.xyzzy.com
DocumentRoot /opt/rt3010/share/html
                                                                            
PerlModule Apache::DBI
                                                                            
PerlRequire /opt/rt3010/bin/webmux.pl
                                                                            
<Location />
    SetHandler perl-script
    PerlHandler RT::Mason
</Location>
                                                                            
ErrorLog /usr/local/www/logs/error_log
TransferLog /usr/local/www/logs/access_log
                                                                            
</VirtualHost>

Database configuration:

postgresql-7.3.4-11
postgresql-server-7.3.4-11
postgresql-libs-7.3.4-11
postgresql-devel-7.3.4-11
postgresql-odbc-7.3-4

There is NO load on the database. A listing of active queries
doesn’t even begin to show any long running transactions. So
please, don’t tell me to use mysql because it’s faster. There’s
no database load at all, it’s the stuff in Apache that’s slow.

Let me know how I can help make RT faster. Two 1GHZ processors should be
plenty of horse power to run a system such as this.

Regards,

dan

RT-Users mailing list
RT-Users@lists.bestpractical.com
The rt-users Archives