Sometimes the Help text on /metrics
endpoint just isn't enough to explain what a certain metric means. This document's objective is to give further explanation of KubeVirt related metrics.
Kubevirt's version information
Labels:
goversion
- GO version used to compile this version of KubeVirtkubeversion
- Git commit refspec that created this version of KubeVirt
The total amount of VMIs per node and phase.
Labels:
phase
- Phase of the VMI. It can be one of Virtual Machine Instance Phasesnode
- Node where the VMI is running on.
All VMI metrics listed below contain, but are not limited to, these three labels for identifying purposes:
name
- VMI's name given on its specification.namespace
- Namespace which the given VMI is related to.node
- Node where the VMI is running on.
Total resident memory of the process running the VMI.
The total amount of usable memory.
The total amount of unused memory as seen by the domain.
The amount of traffic that is being read and written in swap memory.
Extra labels:
type
- Whether the data is being transmitted or received.in
when transmitting andout
when receiving.
Counter of network errors when transmitting and receiving data.
Extra labels:
interface
- Which network interface that errors are occurring.type
- Whether the error occurred when transmitting or receiving data.tx
when transmitting andrx
when receiving.
The total amount of traffic that is being transmitted and received.
Extra labels:
interface
- Which network interface that errors are occurring.type
- Whether the error occurred when transmitting or receiving data.tx
when transmitting andrx
when receiving.
The total amount of packets that are being transmitted and received.
Extra labels:
interface
- Which network interface that errors are occurring.type
- Whether the error occurred when transmitting or receiving data.tx
when transmitting andrx
when receiving.
Counter of read and write operations per disk device.
Extra labels:
drive
- Disk device that is being written/read.type
- Whether it's a read or write operation.
Total time spent on read and write operations per disk device.
Extra labels:
drive
- Disk device that is being written/read.type
- Whether it's a read or write operation.
The total amount of data read and written per disk device.
Extra labels:
drive
- Disk device that is being written/read.type
- Whether it's a read or write operation.
The total amount of time spent in each vcpu state
Extra labels:
id
- Identifier to a single Virtual CPU.state
- Identify the Virtual CPU state. It can be one of libvirt vcpu's states:OFFLINE
,RUNNING
orBLOCKED
Improving Kubevirt's Observability is a important topic and we are currently working on new metrics.
A design proposal and its implementation history can be seen here