cpu.util multi processors

[Raph]

I am currently running zabbix (1.1.4) on centos 4.4, it has one HT xeon processor (2 'virtual' cpu's).

My issue here is that i cannot for the life of me display specific cpu utilization based on CPU. Within linux, it shows two cpus so I know it's the issue is not with the OS, but somewhere in my configuration for zabbix..

This works fine now.
system.cpu.util[all,,avg1]

Anything else fails, ie:
system.cpu.util[1]
system.cpu.util[0]
system.cpu.util[cpu0,,avg1]
system.cpu.util[0,user,avg1]
and so on, as soon as I specify a CPU # nothing is reported from the agent.

Not sure if this helps, but:
#bash: ./zabbix_agentd -c /etc/zabbix/zabbix_agentd.conf -p |grep cpu
cpu[idle1] [m|ZBX_NOTSUPPORTED]
system.cpu.switches [m|ZBX_NOTSUPPORTED]
system.cpu.intr [u|419879155]
system.cpu.util[all,user,avg1] [u|14]
system.cpu.load[all,avg1] [d|0.500000]

I'm kind of lost here, i'm not sure if I need to specify CPU's somewhere in the configuration for the agent, or if my syntax is incorrect, or what.. Any help would be MUCH appreciated

Thanks in advanced!
Raph

[Raph]

nothing?

[phork]

I'm assuming you left your datatype as an int, try changing that to a float and see if that works.

[Raph]

I'm assuming you left your datatype as an int, try changing that to a float and see if that works.

I wish that was it, but it is indeed set to float.

thanks for the reply tho!

i mean is there a place I can see if zabbix even sees more than one cpu period? is there an agent specific setting perhaps?

[phork]

I check cpu utilization on quad logical processor systems and it works without issue (system.cpu.util[,,avg1] specifically). I assuming you have this setup as an agent check (I don't know if it will let you add it as a simple/aggregate or whatever else), so one other thing to check is that the user the zabbix agent runs as has permissions to access the proc filesystem. This is the default on every distro I know of, but if you added additional security products it is possible the user can't grab the info they need. If you can cat /proc/cpuinfo you should be fine.

[phork]

I check cpu utilization on quad logical processor systems and it works without issue (system.cpu.util[,,avg1] specifically). I assuming you have this setup as an agent check (I don't know if it will let you add it as a simple/aggregate or whatever else), so one other thing to check is that the user the zabbix agent runs as has permissions to access the proc filesystem. This is the default on every distro I know of, but if you added additional security products it is possible the user can't grab the info they need. If you can cat /proc/cpuinfo you should be fine.

[Raph]

I check cpu utilization on quad logical processor systems and it works without issue (system.cpu.util[,,avg1] specifically). I assuming you have this setup as an agent check (I don't know if it will let you add it as a simple/aggregate or whatever else), so one other thing to check is that the user the zabbix agent runs as has permissions to access the proc filesystem. This is the default on every distro I know of, but if you added additional security products it is possible the user can't grab the info they need. If you can cat /proc/cpuinfo you should be fine.

i can cat /proc/cpuinfo as the user 'zabbix' on the system (it displays 2 cpus)

Just to be clear here, [system.cpu.util[,,avg1] works fine, but I dont want an average of both 'cpus'. I'd like to see individual usage per cpu.. As soon as I attempt to specify a cpu, that's when it stops reporting. I am using the zabbix_agentd for reporting.

[phork]

Oh, I didn't realize it worked when you didn't specifiy a CPU identifier. I'm not sure the mechanisam zabbix uses to calculate this, but my guess would be something with it being two logical CPUs and now two physical CPUs. Also, you really shouldn't need to check each logical CPU's load as in just about any circumstance each "CPU" should have a similar load (unless your changing the affinity of processes and such). Just my $0.02.