I'm using system.cpu.total.norm.pct for average CPU usage but in some of the servers the reading is incorrect, which I guess is probably related to incorrect reading of system.cpu.cores
CPU usage reported by system.cpu.total.norm.pct is nearly double the value reported by performance counter or task manager.
Task Manager Data:
Sockets: 4
Cores: 64
Logical processors: 128
Metricbeat Reading:
system.cpu.cores: 32
Spec:
version: 7.2
Processor: Intel(R) Xeon(R) CPU E7-4850 v4 @2.10 Ghz (4 processors)
Windows: Windows Server 2016 Standard
Windows Build: 14393.3085
It's worth mentioning reading from windows module with performance counters is accurate.
- instance_label: processor.name
instance_name: Total
measurement_label: processor.time.total.pct
query: '\Processor Information(_Total)\% Processor Time'