Windows Cluster Support for Monitoring V2


Available as of v2.5.8

Starting at Monitoring V2 14.5.100 (used by default in Rancher 2.5.8), Monitoring V2 can now be deployed on a Windows cluster and will scrape metrics from Windows nodes using prometheus-community/windows_exporter (previously named wmi_exporter).

Comparison to Monitoring V1

Unlike Monitoring V1 for Windows, metrics collected by windows_exporter will be labeled as windows_ instead of wmi_ in accordance to a naming change from upstream from wmi_exporter to windows_exporter.

In addition, Monitoring V2 for Windows will no longer require users to keep port 9796 open on Windows hosts since the host metrics will published directly onto a port exposed on the windows-exporter Pod. This feature was powered by recent changes made by wins v0.1.0 to support publishing ports exposed on the hostNetwork on Pods that use wins to run a privileged Windows binary as a host process.

Cluster Requirements

Monitoring V2 for Windows can only scrape metrics from Windows hosts that have a minimum wins version of v0.1.0. To be able to fully deploy Monitoring V2 for Windows, all of your hosts must meet this requirement.

If you provision a fresh RKE1 cluster in Rancher 2.5.8, your cluster should already meet this requirement.

Upgrading Existing Clusters to wins v0.1.0

If the cluster was provisioned before Rancher 2.5.8 (even if the current Rancher version is 2.5.8), you will not be able to successfully deploy Monitoring V2 for Windows until you upgrade the wins version on each host to at least v0.1.0.

To facilitate this upgrade, Rancher 2.5.8 has released a brand new Helm chart called rancher-wins-upgrader.

Prerequisite: Make sure Monitoring V1 for Windows is uninstalled.

  1. Deploy rancher-wins-upgrader with the following override:

    1. # Masquerading bootstraps the wins-upgrader installation via
    2. # a previously whitelisted process path since the normal install path,
    3. # c:\etc\rancher\wins\wins-upgrade.exe is not normally whitelisted.
    4. # In this case, we are using the previously whitelisted process
    5. # path used by Monitoring V1.
    6. masquerade:
    7. enabled: true
    8. as: c:\\etc\wmi-exporter\wmi-exporter.exe

    Note for Non-Default Windows Prefix Path: If you set up the RKE cluster with a cluster.yml that has a non-default win_prefix_path, you will need to update the masquerade.as field with your prefix path in place of c:\\.

    For example, if you have win_prefix_path: 'c:\host\opt\', then you will need to set as: c:\host\opt\etc\wmi-exporter\wmi-exporter.exe.

  2. Once all your hosts have been successfully upgraded, please ensure that you deploy the Helm chart once again with default values to avoid conflicts with the following settings:

    1. masquerade:
    2. enabled: false

Result: The hosts are ready for Monitoring V2 to be installed. You may choose to uninstall the rancher-wins-upgrader chart or keep it in your cluster to facilitate future upgrades.

For more information on how it can be used, please see the README.md of the chart.