CPU Load question

Hello All,

How can we relate the CPU graph on the dashboard to “over usage” ?
We have about 15-20 concurrent calls and we are seeing CPU spikes of around 1.6 & 1.8 via the dashboard management.

Should we be bumping up CPU for something like this?


I find the “network usage” graph to better represent “over use”.

Having said that, how would you rate 1.6, 1.8 CPU load? OK or High?

If you have 1 CPU, it’s high. If you have 24, it’s low.

LOL… OK so with 4 CPU’s do you feel like this number is OK, or should we bump up the horsepower?

You’re asking ‘how long is a piece of string’, unfortunately. Why is the load high? Is it Disk IO bound? CPU Bound? You’ll need to investigate your performance issues yourself, and make that call.

The commands ‘top’ and ‘sar -d 5 999’ will be helpful in this. (You can install sar with ‘yum install sysstat’, if you don’t have it installed allready)

About hardware resources in general, not CPU but RAM :

this is the partial output of TOP when sorted by memory usage (just a couple of processes that occupy the most of memory: )

Mem:   1021656k total,   958280k used,    63376k free,    94456k buffers
Swap:  1303408k total,   104640k used,  1198768k free,   380940k cached
PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+   COMMAND                                                            
29745 asterisk  20   0 95172  37m 5128 S  0.0  3.7   0:02.45 httpd                                                               
29748 asterisk  20   0 95172  36m 5176 S  0.0  3.7   0:01.89 httpd                                                               
29746 asterisk  20   0 95172  36m 4628 S  0.0  3.7   0:02.25 httpd                                                               
29741 asterisk  20   0 91076  34m 5096 S  0.0  3.4   0:02.97 httpd                                                               
29750 asterisk  20   0 91076  33m 5232 S  0.0  3.4   0:01.71 httpd                                                               
8204 asterisk  20   0 91076  33m 4672 S  0.0  3.3   0:01.89 httpd                                                               
29747 asterisk  20   0 91076  33m 4692 S  0.0  3.3   0:01.90 httpd  

I noticed, that memory usage by these httpd processes seems to be climbing.
To the point that they seem to occupy ~7% of memory EACH, and preventing correct operation of e.g. the update process (via web page or via fwconsole or distro update scripts). I don’t have the exact record, but top was showing buffers really low and almost no free memory. Like all those httpd processes (and if I toggle command line it’s not helpful either, just shows /usr/sbin/httpd) were crammed into RAM. The only way to make them free the memory, is to service httpd restart. Yes my system only has THIS much of total RAM… so I learned to periodically check top output, and do service httpd restart before running the updates.
Anything you think this could be related to ? FreePBX 13.0.189 but had this effect for at least 2 or 3 updates back…