monitoring SMP system.cpu.util[all,user,avg1]

[netod]

On SMP systems, is it possible to monitor individual CPUs separately rather than having average CPU utilisation for both?

[Calimero]

Documentation says first parameter is CPU number. Maybe something like this works:
system.cpu.load[0,avg1]

Can't test it because zabbix is deployer yet on production servers...

[netod]

Documentation says first parameter is CPU number. Maybe something like this works:
system.cpu.load[0,avg1]

Can't test it because zabbix is deployer yet on production servers...

THanks for your reply.

The item in question is actually util not load but the same arguments apply I guess..

system.cpu.util[0,,avg1]
and
system.cpu.util[1,,avg1]

is what I used and gives me Not supported
zabbix agent 1.1.3

[mtggit]

Same issue

I use zabbix on my dual core machine running Suse Linux 9 server, which also give me Not Supported. I use the same parameter of:

system.cpu.util[0,user,avg1]
system.cpu.util[1,user,avg1]

Is there something wrong with the zabbix agentd?

The /proc/stat result is as follows:

cpu 1752813 28178 13798483 157916457 477931 3916 17293
cpu0 374182 5789 3075588 39942041 97068 811 3329
cpu1 474207 18731 3747627 38951247 290386 3057 13496
cpu2 483194 2747 3624762 39343125 44710 24 190
cpu3 421230 909 3350504 39680043 45765 22 277
intr 2234879617 434995466 0 0 0 5 0 0 0 1778516163 0 0 0 0 0 0 0 688216 19493295 25 0 12 1178157 8278 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
ctxt 11132943968
btime 1179968817
processes 137451
procs_running 3
procs_blocked 0

[Andrew]

I use zabbix on my dual core machine running Suse Linux 9 server, which also give me Not Supported. I use the same parameter of:

system.cpu.util[0,user,avg1]
system.cpu.util[1,user,avg1]

May be try 1.4.2?

I see in release notes:

What's New in 1.4.2

:: Per-CPU statistics

Support of per-CPU system.cpu.util and support of system.cpu.num under Linux and Solaris.