system.cpu.util Type of received value [5.5] is not suitable

[eger]

I am receiving the following errors for one of my agents:

12692:20100330:123534.833 Item [nyc01.mydomain.com.com:system.cpu.util[,,avg5]] error: Type of received value [5.5] is not suitable for value type [Numeric (integer 64bit)]

12696:20100330:123545.566 Item nyc01.mydomain.com.com:system.cpu.util[,,avg1]] error: Type of received value [5.216667] is not suitable for value type [Numeric (integer 64bit)]

The machine is Windows 2003. However this item seems to be working fine on all my other Windows 2003 machines. The value received back looks liek a correct decimal.

Any ideas?

[rincewind]

I am receiving the following errors for one of my agents:

12692:20100330:123534.833 Item [nyc01.mydomain.com.com:system.cpu.util[,,avg5]] error: Type of received value [5.5] is not suitable for value type [Numeric (integer 64bit)]

12696:20100330:123545.566 Item nyc01.mydomain.com.com:system.cpu.util[,,avg1]] error: Type of received value [5.216667] is not suitable for value type [Numeric (integer 64bit)]

The machine is Windows 2003. However this item seems to be working fine on all my other Windows 2003 machines. The value received back looks liek a correct decimal.

Any ideas?

May be try to change data type to Numeric(float)?
Also, I highly recommend to use perf_counter[\Processor(_Total)\% Processor Time] on windows servers to monitor cpu load instead of syste.cpu.util. This way you get exacly what you can see in the taskmanager (in percent) of certain server and even to monitor each cpu core separately.

[eger]

Changing them to float got them working.

I also added the perf_counter to compare. What is strange, is that the perf_counter is reporting different values than system.cpu.util. I wonder which one is less overhead?

[rincewind]

I also added the perf_counter to compare. What is strange, is that the perf_counter is reporting different values than system.cpu.util. I wonder which one is less overhead?

I do not have any idea how system.cpu.util works, but I can describe how perf_counter does. It just simple get perflib counters for target system (by the way, windows task manager uses this library) and get proper values. So, if we request "\Processor(_Total)\% Processor Time", we'll get average load in percent of all processors. Can be modify to "\Processor(0)\% Processor Time" to monitor just only CPU0 processor load. Also, you can see and manually select proper performance counters by standard windows utility perfmon. You can even handle cpu/memory/disk utilization per-application. For example, I monitor overall SQL server memory allocation by perf_counter["\SQLServer:Memory Manager\Total Server Memory (KB)"] and sql server load in transactions per second by perf_counter["\SQLServeratabases(_Total)\Transactions/sec"].

[eger]

Thanks. I have used the performance counters before. I may start a new thread on the differences of the zabbix CPU util and the perf counter. I would have expected the 1min CPU util key to be the same or similar to the perf counter...