per CPU monitoring on Linux ... max 10 CPU?

[username_taken]

Just put in a new 16core linux (centos) server to replace a underpowered 8 core server. Using system.cpu.util to monitor the load on each CPU ( system.cpu.util[1,system,avg1], etc ) and it appears that 10 CPU's is some sort of limit for zabbix to report back, as 11 to 16 get a response of 0 ( not unsupported, just zero ).

Some custom tests I have written show processes running against those CPUs so there should be something showing. I'm wondering if the zabbix source code only supports a single digit integer for CPU testing? Remembering that Zabbix translates CPU from the server as n+1 ( i.e. CPU1 in zabbix is really CPU0 on server ) this would make sense for it only reporting 10 cpus ( 0..9 ).

Anyone else experience this, or know how to work around it ?

[richlv]

cpu numbering from 0 is a linux thing.
if you can verify that with all cpus loaded zabbix still shows 0 for cpus past 9/10 (maybe something like for i in $(seq 1 16); do (cat /dev/urandom | md5sum)&; done ), please report that on the tracker

[username_taken]

Yup I understand that CPU numbering from 0 is a Linux thing, and zabbix uses 'n-1' to translate.

Definately not working, I don't need to stress the CPUs as I can see activity on CPUS > 10 ... plus idle time is also coming back as 0. it should be 100% if there's no cpu used

I'll add this to tracker.

[username_taken]

added to tracker

https://support.zabbix.com/browse/ZBX-1184