vicidial.org

Good day guys and gals

I would appreciate a bit of assistance in tuning a heavily loaded express install but first here is the customer's setup:

Version: 2.12b0.5
BUILD: 160122-1401
SVN Version: 2483
DB Schema Version:1452
Kernel version - 4.1.39-56-default
Express Install

Dell R510
2X Intel(R) Xeon(R) CPU X5660 @ 2.80GHz (6 Cores 12 threads)
64 GB ECC RAM
2 x 1TB Enterprise SSD drives (RAID 1)
Perc H700 RAID controller (I know, I know it's not the LSI MegaRaid but this will get changed in the new cluster build we are planning )

Ok so my issue here is the following:

One server 88 agents (I know right!!!!!) Issue is that Vicidial is randomly kicking agents out of their sessions (Obviously)

I just need to keep this dialer running as is until we can complete their cluster install with new hardware and done the correct way ofcourse and not the way the previous service provider did things!

So far what I have done this evening (South Africa time ) is the following:

Copied /usr/src/astguiclient/conf/server-tuning.conf to /etc/apache2/server-tuning.conf

Enabled opcache under /etc/php5/apache2/php.ini in the [opcache] section and set all the correct options and values as per below:

opcache.enable=1
opcache.memory_consumption=1024
opcache.interned_strings_buffer=4
opcache.max_accelerated_files=2000
opcache.max_wasted_percentage=5
opcache.validate_timestamps=1
opcache.revalidate_freq=10
opcache.error_log=/var/log/apache2/opcache.log
Modifed /etc/my.cnf to allow 2048 connect, 4GB key buffer, 0 query cache, 2048 table defs, 8192 file defsI

I have not yet tested the system under load since making the above changes so I guess I will find out tomorrow but I just want to know if I missed anything?

I would greatly appreciate your guys help!

Thank you

Ouch!

If possible, you should also archive the logs as frequently as the system will allow during downtime.

Thank you Matt, I take it what I did above was fine then?

I will try and cram as much archiving of logs in as I can.

Not really sure what effect your changes will have, all bets are off when you're dealing with an overloaded system

Just a quick update. After the changes I made as per above, we benchmarked the system to see if we could get it to fall over by loading more agents on as the system was dialing without a problem. (Full blended, dial level 2 on most campaigns)
We added all the users (extra users) and the system still behaved.

The GUI was also at least 50% faster than before so the changes made most definitely had a massive impact.

This will now at least keep the customer going until we can sort out the commercials with them to get the hardware for their new cluster install.

I know not ideal but we bought ourselves some time for what it's worth.

Ciao