1. Documentation
×

Machine Config Operator

The Machine Config Operator provides useful information to cluster administrators and controls what is running directly on the bare-metal host.

The Machine Config Operator differentiates between different groups of nodes in the cluster, allowing control plane nodes and worker nodes to run with different configurations. These groups of nodes run worker or application pods, which are called MachineConfigPool (mcp) groups. The same machine config is applied on all nodes or only on one MCP in the cluster.

For more information about how and why to apply MCPs in a telco core cluster, see Applying MachineConfigPool labels to nodes before the update.

For more information about the Machine Config Operator, see Machine Config Operator.

Purpose of the Machine Config Operator

The Machine Config Operator (MCO) manages and applies configuration and updates of Fedora CoreOS (FCOS) and container runtime, including everything between the kernel and kubelet. Managing FCOS is important since most telecommunications companies run on bare-metal hardware and use some sort of hardware accelerator or kernel modification. Applying machine configuration to FCOS manually can cause problems because the MCO monitors each node and what is applied to it.

You must consider these minor components and how the MCO can help you manage your clusters effectively.

You must use the the MCO to perform all changes on worker or control plane nodes. Do not manually make changes to FCOS or node files.

Applying several machine config files at the same time

When you need to change the machine config for a group of nodes in the cluster, also known as machine config pools (MCPs), sometimes the changes must be applied with several different machine config files. The nodes need to restart for the machine config file to be applied. After each machine config file is applied to the cluster, all nodes restart that are affected by the machine config file.

To prevent the nodes from restarting for each machine config file, you can apply all of the changes at the same time by pausing each MCP that is updated by the new machine config file.

Procedure

  1. Pause the affected MCP by running the following command:

    $ oc patch mcp/<mcp_name> --type merge --patch '{"spec":{"paused":true}}'

  2. After you apply all machine config changes to the cluster, run the following command:

    $ oc patch mcp/<mcp_name> --type merge --patch '{"spec":{"paused":false}}'

This allows the nodes in your MCP to reboot into the new configurations.