Help, open and close ticket is very slow.

[mikeoldyang]

We are recently installing a otrs system in our company. Every thing is OK but the performance is very slow when we open and close a ticket. Normally we need more than 10 seconds to close a ticket in web, sometimes is 20 seconds. And need more than 5 seconds to open a ticke in web.

We have tried to add CPU/Memory(Now is 4cores and 8G), active mod_perl, archive ticket, enable staticDB, disable animation and something else base on http://otrs.github.io/doc/manual/admin/ ... uning.html

but no luck with any improvement.

I want to know how about your status? And any suggestion?

Thanks

Mike

FYI:
The environment is:
------------------------------------------------------------------------------------------------------------------------------------------
DOCUMENT_ROOT /var/www/html
GATEWAY_INTERFACE CGI/1.1
HTTP_CONNECTION Keep-Alive
HTTP_DNT 1
HTTP_HOST xxx
MOD_PERL mod_perl/2.0.4
MOD_PERL_API_VERSION 2
PATH /sbin:/usr/sbin:/bin:/usr/bin
REMOTE_ADDR xxx
REMOTE_PORT xxx
SCRIPT_FILENAME /opt/otrs/bin/cgi-bin/index.pl
SERVER_ADDR xxx
SERVER_ADMIN root@localhost
SERVER_NAME xxx
SERVER_PORT 80
SERVER_PROTOCOL HTTP/1.1
SERVER_SIGNATURE <address>Apache/2.2.15 (CentOS) Server at xxx Port 80</address>
SERVER_SOFTWARE Apache/2.2.15 (CentOS)

Agents 249
Articles 11704
Articles per ticket (avg) 3.04
Attachment size (avg) 134.00 KB
Attachments per ticket (avg) 0.24
Customers with at least one ticket 849
Groups 127
Install directory /opt/otrs
Months between first and last ticket 1.89
Operating system linux
Product OTRS 3.3.4
Roles 104
Tickets 3845
Tickets per month (avg)
------------------------------------------------------------------------------------------------------------------------------------------

[Lubomirsb]

Restart otrs service ?

[mikeoldyang]

Restart otrs service ?

have restarted many times...

[alexus]

Try to switch off any notifications or just disable outgoing email. Do you have any event-based developments for your OTRS?

[mikeoldyang]

Try to switch off any notifications or just disable outgoing email. Do you have any event-based developments for your OTRS?

Thanks alexus. I set 8 notification(events) and 11 geneal agent which are event-based both.

Also I noticed that the notification email has been sent but the web is still in freshing. I mean when I close a ticket which customer is me, I can get notification email (close ticket) within 2 seconds but the page in web need 20 seconds to close.

Any suggestion?

Mike

[reneeb]

Is something else running on the machine? Did you check the logs if anything suspicious popped up?

[mikeoldyang]

Is something else running on the machine? Did you check the logs if anything suspicious popped up?

The server is only for otrs application. I run Support Assessment and solved some problem, like the time difference between application server time and database server time, and active mod_perl. But the system is still slow. There is only one warning now is "Packages not correctly installed: FAQ 2.3.1; ITSMConfigurationManagement 3.3.4; ITSMIncidentProblemManagement 3.3.4; Survey 2.3.2; ." I installed a new otrs in another machine, the warning is there also but no performance issue.

[reneeb]

I installed a new otrs in another machine, the warning is there also but no performance issue.

Then check what the differences between the servers are.

[mikeoldyang]

I installed a new otrs in another machine, the warning is there also but no performance issue.

Then check what the differences between the servers are.

Actually there is no difference, everything is same. Same sysconfig.pm, the DB is a backup copy from the slow system.

[schulmann]

You can try this: viewtopic.php?f=62&t=23757&hilit=+keepalive#p94002

[mejiaufv]

It really does sound like an issue with not enough RAM being available. I had a similar problem a few weeks back and the info on the thread that schulmann linked to solved it for me.

[mikeoldyang]

It really does sound like an issue with not enough RAM being available. I had a similar problem a few weeks back and the info on the thread that schulmann linked to solved it for me.

Thanks for your suggestion, but now my machine has already 8G memory.

[mikeoldyang]

You can try this: viewtopic.php?f=62&t=23757&hilit=+keepalive#p94002

Thanks. I saw that document, and it seems have two questions in one topic. I got you mentioned that:
------------------
We have specified a rule of thumb for OTRS on Linux with Apache:
•KeepAlive Off
•5 Apache processes (MaxClients)
•For every 5 concurrent agents 1 additional Apache process
•For every Apache process 250 MB of RAM
------------------

so is it your suggesion? and now the machine has already 8GB memory. is it enough? or it should do some configration?

Mike

[crythias]

Check the performance Log option and Support Assessment.

[Giulio Soleni]

is your database MySQL with InnoDB engine?
I noticed a small performance inproovement when setting

innodb_flush_log_at_trx_commit = 2

on /etc/my.cnf

The drawback of this setting is to allow a few seconds of data loss in case of power failure... If you are not so concerned with that, you may try to set this parameter, restart MySQL and check from the benchmarks if the performances have been improoved a little bit.

There is also a perl script made by Major Haydenthat may be useful for the performance tuning of MySQL database.
Check it out on mysqltuner.com
Of course it has not been developed to substitute a dba, but it may be useful to give you some possible hints... you may run it safely on your current database implementation and then decide to apply or not the suggested configurations.

HTH
Giulio

[schulmann]

and now the machine has already 8GB memory. is it enough?

You didn't mention the number of concurrent (simultaneously logged in) agents.
My suggestion hasn't been tested with thousands of service categories or other exotic settings.
Therefore I don't know if 8 GB of RAM are enough for your environment.

Additionally, on linux "vmstat 1" may provide some useful information.

You can try my suggestion. If it is not successful you can simply restore your previous configuration.

[mikeoldyang]

and now the machine has already 8GB memory. is it enough?

You didn't mention the number of concurrent (simultaneously logged in) agents.
My suggestion hasn't been tested with thousands of service categories or other exotic settings.
Therefore I don't know if 8 GB of RAM are enough for your environment.

Additionally, on linux "vmstat 1" may provide some useful information.

You can try my suggestion. If it is not successful you can simply restore your previous configuration.

Thanks schulmann. Normally we have about 10 agents online at same time. I run the vmstat 1 as below, please help to have a look.

3.png

[schulmann]

I run the vmstat 1 as below, please help to have a look.

What does "vmstat 1" say if you set KeepAlive to Off and MaxClients to 10?

[mikeoldyang]

Hi everybody, I finally got the root cause. It's SMTP.

When I pointed the smtp server to a wrong host, the speed of opening and closing ticket is very fast(but customers and agents are not able to receive mail from otrs). Once I set the smtp server to a right server, the closing and opening will be very slow.

But the SMTP(Microsoft Exchange Server) is no problem. I saw there was a defaul option called "sendmail", did you use it? How to use it?

33.png