System Load Average

[kolucoms6]

My server shows : :

9 agents logged in on all servers
System Load Average: 10.49


I have 1 mb lease line.

Why system load average is so HIGH ?

Due to this, agent cant hear the customer

[ykhan]

Some factors affecting Load Avg:

1. Codecs being used. e.g. Agent phone = ULAW, Provider = G729 will result in higher transcoding and therefore increase server load. Basically the higher transcoding = higher load
2. Number of calls being dialed per agent. More calls = More load
3. More Call recording = More Load
4. Linux Kernel: If you rebuilt the kernel and did not take out all unnecessary modules, then you may have additional modules compiled or loaded that are not necessary and adding load on the server.
5. Server hardware: All of the above correspond to the type of hardware you have. If you are trying to run 9 agents on a P4, you will probably have issues, but same on a Quad Core machine will be much better performance.

These are just a few of the factors that could be influencing your server load. Take a look at them and try to test out the best setup for you. And remember if you are using your bandwidth for only calls, you can expect the following capacity breaks:

ULAW 90 kbps = 11 simultaneous lines per MB
GSM 30 kbps = 33 simultaneous lines per MB
G729 13 kbps = 77 simultaneous lines per MB

I hope this helps you.

[kolucoms6]

1. Codecs being used. e.g. Agent phone = ULAW, Provider = G729 will result in higher transcoding and therefore increase server load. Basically the higher transcoding = higher load : We use XLITE and on server, its G729.Provider says he supports both.
2. Number of calls being dialed per agent. More calls = More load : 1:2.5
3. More Call recording = More Load : 100% call recording
4. Linux Kernel: If you rebuilt the kernel and did not take out all unnecessary modules, then you may have additional modules compiled or loaded that are not necessary and adding load on the server. : NO IDEA
5. Server hardware: All of the above correspond to the type of hardware you have. If you are trying to run 9 agents on a P4, you will probably have issues, but same on a Quad Core machine will be much better performance. :

Processors 2
Model Intel(R) Xeon(TM) CPU 3.00GHz
CPU Speed 2.99 GHz
Cache Size 2.00 MB
System Bogomips 11971.7






Memory Consumption is 95%, which I guess is causing issues. Though I have 4 GB RAM DDR2 and only 9 agents logged in.

[ykhan]

Try the following:

1. Turn of call recording (just to test) and see if it lowers your Load Avg.
2. Check the load avg. before you start dialing to get a base line of your resource usage. I have noticed that after some time the RAM is eaten up so make sure you reboot the system once every two days or so.

Another thing that can affect load is the type of hard disk, SATA/SCSI is higher performance than IDE. Also, check for total bandwidth usage with "iftop" at the console. If you do not have the g729 codec installed on your server then your system won't be using it. Post your sip.conf entry for the provider and an agent phone.

If you still have questions, call me 647-891-5426, it just works better for trouble shooting.

[kolucoms6]

Try the following:

1. Turn of call recording (just to test) and see if it lowers your Load Avg. : I tried it and Now I see LoadAvg is OK NOW[/color]
2. Check the load avg. before you start dialing to get a base line of your resource usage. I have noticed that after some time the RAM is eaten up so make sure you reboot the system once every two days or so.
We Restart the server daily.
Another thing that can affect load is the type of hard disk, SATA/SCSI is higher performance than IDE. Also, check for total bandwidth usage with "iftop" at the console. If you do not have the g729 codec installed on your server then your system won't be using it. Post your sip.conf entry for the provider and an agent phone. We have SATA.

If you still have questions, call me 647-891-5426, it just works better for trouble shooting.

[kolucoms6]

All recording calls can create Issue ?

[ykhan]

Yes, recoding calls places a fair bit of burden on the server resources. You can do two things:

1. Add a recording delay. Normally if its a Answering machine or not interested the call will be hung up in 10-15 seconds. So it would be safe to use recording delay of abou that much time.
2. Change mix time for the recordings in your cron to more than 5 minutes. Instead of every 5 minutes, try changing to once every 8 hours (shift).

[kolucoms6]

1) Where to change Recording delay ?

2) What is "mix time for the recordings" and how to change it ?

[ykhan]

Recoding delay option is on the campaign page.

For Recording MIX you have to go into the cron (Linux task scheduler). Type crontab -e at the console.

[kolucoms6]

Changed Recording delay to 20. it was ZERO


crontab -e shows :
______________________________________


[root@vicidialnow ~]# crontab -e
### recording mixing/compressing/ftping scripts
0,3,6,9,12,15,18,21,24,27,30,33,36,39,42,45,48,51,54,57 * * * * /usr/share/astguiclient/AST_CRON_audio_1_move_mix.pl
#0,3,6,9,12,15,18,21,24,27,30,33,36,39,42,45,48,51,54,57 * * * * /usr/share/astguiclient/AST_CRON_audio_1_move_VDonly.pl
1,4,7,10,13,16,19,22,25,28,31,34,37,40,43,46,49,52,55,58 * * * * /usr/share/astguiclient/AST_CRON_audio_2_compress.pl --GSM
#2,5,8,11,14,17,20,23,26,29,32,35,38,41,44,47,50,53,56,59 * * * * /usr/share/astguiclient/AST_CRON_audio_3_ftp.pl --GSM

[kolucoms6]

How to change Recording Mix ?

[kolucoms6]

[root@vicidialnow ~]# df -h
Filesystem Size Used Avail Use% Mounted on
/dev/sda2 288G 196G 78G 72% /
/dev/sda1 99M 11M 83M 12% /boot
tmpfs 1.5G 0 1.5G 0% /dev/shm

Can space be a problem ?

If yes, then I can delete recording to clear the HDD.

Kindly advice.

[williamconley]

i'm not sure about defrag on a linux box, but you should also turn off logging unless you are troubleshooting (which is in the server settings under ADMIN, and in "logger.conf").

[kolucoms6]

I below wil help me . Let me try it.
_____________________________________-

System Performance - Setting this option to Y will enable logging of system performance stats for the server machine including system load, system processes and Asterisk channels in use. Default is N.

Server Logs - Setting this option to Y will enable logging of all VICIDIAL related scripts to their text log files. Setting this to N will stop writing logs to files for these processes, also the screen logging of asterisk will be disabled if this is set to N when Asterisk is started. Default is Y.

[kolucoms6]

Instead of every 5 minutes, try changing to once every 8 hours (shift).

In which file and what do I have to change ?

[okli]

You may add this to your cron file (crontab -e to edit it) for quick reference:

Code: Select all

#Minute (0-59),
#|      Hour (0-23),
#|      |       Day of the month (1-31),
#|      |       |       Month of the year (1-12),
#|      |       |       |       Day of the week (0-6 with 0=Sunday).
#|      |       |       |       |       Commands
#|      |       |       |       |       |
#----------------------------------------------------------
############################################

Read this:
http://webmasterworkshop.com/guides/crontab_guide.shtml

In short- you have to edit the entries you posted above, understanding what each element of crontab means. Should be more helpful than if someone post you how exactly to be changed.

### recording mixing/compressing/ftping scripts
0,3,6,9,12,15,18,21,24,27,30,33,36,39,42,45,48,51,54,57 * * * * /usr/share/astguiclient/AST_CRON_audio_1_move_mix.pl
#0,3,6,9,12,15,18,21,24,27,30,33,36,39,42,45,48,51,54,57 * * * * /usr/share/astguiclient/AST_CRON_audio_1_move_VDonly.pl
1,4,7,10,13,16,19,22,25,28,31,34,37,40,43,46,49,52,55,58 * * * * /usr/share/astguiclient/AST_CRON_audio_2_compress.pl --GSM
#2,5,8,11,14,17,20,23,26,29,32,35,38,41,44,47,50,53,56,59 * * * * /usr/share/astguiclient/AST_CRON_audio_3_ftp.pl --GSM

[Baylink]

Most Linux filesystems do not require defragmentation; their allocation policies are designed not to *cause* much of it, in the first place. You can force fragmentation in pathological cases, but...

The traditional way to defrag a *nix file system is to dump and reload it, but that's not really practical.