Page 1 of 1
Load issues - Resumed
Posted:
Wed Aug 02, 2006 1:54 pmby enjay
Still trying to isolate these load issues..
I have 12 agents
6 dialing out 2 LD T1's - Dial Level of 6 - 36 lines
6 dialing out 2 Local T1's - Dial Level of 6 - 36 lines
72 lines dialing at a time 12 total agents..
Load avg between 9-11
What can I look into?
Posted:
Wed Aug 02, 2006 3:02 pmby mflorell
What is your dial timeout set to?
Are you recording all calls?
just to confirm that apache and MySQL are not running on these servers?
what VOIP codecs are you using?
If you run "top" do you see anything taking up a lot of resources?
how much RAM is free?
How often do you reboot these servers?
Posted:
Wed Aug 02, 2006 3:28 pmby lkeller
Enjay is at lunch, so I will fill in some responses for ya:
What is your dial timeout set to? -- 22 seconds
Are you recording all calls? -- yes all calls
Just to confirm that apache and MySQL are not running on these servers? -- Correct, they are on another server
what VOIP codecs are you using? -- ulaw
If you run "top" do you see anything taking up a lot of resources? -- asterisk (obviously) and lame pops up there occasionally.
How much RAM is free? -- that varies.. usually about 190,000k - 200,000k
How often do you reboot these servers? -- We don't really reboot them...
Posted:
Wed Aug 02, 2006 4:26 pmby mflorell
Why is lame running?
if you are mixing these while you are dialing that will really hurt your performance.
As for rebooting. You might want to reboot at least once a week if not nightly.
Posted:
Wed Aug 02, 2006 5:01 pmby enjay
Since its a new project, the servers have only been up maybe a week.. We will rewrite the mixrecordings script to send stuff over and have lame do the conversion there...
We will submit the new script to mantis when we're done.
-enjay
Posted:
Wed Aug 02, 2006 7:13 pmby enjay
/usr/bin/perl /home/cron/AST_send_action_child.pl --data1=460653
that script also tends to take up between 1.5 -> 4.0 CPU while thats not much sometimes as many as 30 of them are running at a time and yea that adds up. That is typically when we see the huge spike in system util..
Posted:
Wed Aug 02, 2006 9:50 pmby mflorell
One of the goals for 2.0 is creating constantly running child send scripts so that they do not need to spawn and die, spawn and die using a lot more resources on a busy system than a constantly running app.
You also might want to see if using the DBI version of that script will help with resource usage.