lab009.md

Lab 9: Exploring KubeVirt metrics

Using the PromUI

Using your brower, head to your node's hostname on port 30090 to find the PromUI.

Metrics in KubeVirt can be differentiated in two types:

• Metrics for components
  • Go internal metrics
  • API requests
• Metrics for VMs
  • Resident memory in bytes
  • Total network traffic in bytes
  • Total storage IOPs
  • Total storage times in ms
  • Total storage traffic in bytes
  • CPU usage in seconds

Note that metrics are a relatively new feature in KubeVirt and are under development.

Let's query PromUI for the resident memory that our VMs are currently using. The metric's name is kubevirt_vm_memory_resident_bytes:

We could also use any of the following VM metric names:

• kubevirt_vm_network_traffic_bytes_total
• kubevirt_vm_storage_iops_total
• kubevirt_vm_storage_times_ms_total
• kubevirt_vm_storage_traffic_bytes_total

NOTE: it is useful to have a look at the PromQL and its functions, some are helpful to shape the metrics into more useful data.

Now let's check the API metrics using rest_client_requests_total{job="kubevirt-prometheus-metrics"}:

Again, these will tend to be ever-growing numbers, thus PromQL functions are needed. Note the information that Promtheus gives us, like code, Pod, method... we could make use of those to create monitoring dashboards and alterting.

Using Grafana

In the student materials ($HOME/student-materials) you'll find a Grafana dashboard we've prepared for you all, so let's import it and see what it shows, perform the following steps in your own workstation/laptop:

## LAB009

wget https://raw.githubusercontent.com/kubevirt/kubevirt-tutorial/master/administrator/ansible/roles/kubevirt/files/student-materials/KubeVirt-grafana-dashboard.json

Now, point your browser to your node's hostname port 30300. You will be greeted by Grafana's login page. Use the following authentication details:

• User: admin
• Password: kubevirtlab123

To import the dashboard click on the plus symbol, to the left of the page and then Import:

Click on the Upload .json File button, find the dashboard file and import it!

The dashboard will open right away:

The dashboard counts the different return codes from the API client requests we saw earlier on PromUI and graphs CPU, memory and IOPs for all the VMs running on the cluster.

This concludes this lab!

<< Previous: Deploy a multi-homed VM using Multus | README | Next: Conclusion >>