Get cpu consumption from KVM VMs
|Assignee:||Javi Fontan||% Done:|
|Category:||Drivers - Auth|
|Target version:||Release 2.2|
KVM virtual machines lack cpu consumption information
New probes and ganglia readers (from branch feature-446)
poll_xen_kvm.rb - This new probe gets information from XEN and KVM vms.
If it is called with the name of a VM then its monitoring information
is written in standard poll probe way. If called without parameters
a base64-encoded yaml-representation of a hash containing the information
about all the VMs running in the host is returned.
poll_ganglia.rb - This probe gets information from ganglia pushed by gmetric.
Consumes the information generated by poll_xen_kvm.rb
ganglia_probe - Gets host monitoring information from ganglia. It pushes all
metrics that start by OPENNEBULA_
Ganglia.rb - library that holds ganglia reading code
#2 Updated by Shi Jin over 10 years ago
I am very happy to see the monitoring code here.
Just one thing: the poll code calls the ps command for CPU utilization which is the average from the starting of the VM while I would say we should use something like top to get more instantaneous (averaged since last top update, typically 3 seconds) CPU monitoring. Of course we need to first define what the CPU column from the "onevm list" output is for: instantaneous or average from the start.