Machine configuration tasks | Post-installation configuration

About the Machine Config Operator
Using MachineConfig objects to configure nodes
Configuring MCO-related custom resources

There are times when you need to make changes to the operating systems running on OKD nodes. This can include changing settings for network time service, adding kernel arguments, or configuring journaling in a specific way.

Aside from a few specialized features, most changes to operating systems on OKD nodes can be done by creating what are referred to as MachineConfig objects that are managed by the Machine Config Operator.

Tasks in this section describe how to use features of the Machine Config Operator to configure operating system features on OKD nodes.

About the Machine Config Operator

OKD 4.13 integrates both operating system and cluster management. Because the cluster manages its own updates, including updates to Fedora CoreOS (FCOS) on cluster nodes, OKD provides an opinionated lifecycle management experience that simplifies the orchestration of node upgrades.

OKD employs three daemon sets and controllers to simplify node management. These daemon sets orchestrate operating system updates and configuration changes to the hosts by using standard Kubernetes-style constructs. They include:

The machine-config-controller, which coordinates machine upgrades from the control plane. It monitors all of the cluster nodes and orchestrates their configuration updates.
The machine-config-daemon daemon set, which runs on each node in the cluster and updates a machine to configuration as defined by machine config and as instructed by the MachineConfigController. When the node detects a change, it drains off its pods, applies the update, and reboots. These changes come in the form of Ignition configuration files that apply the specified machine configuration and control kubelet configuration. The update itself is delivered in a container. This process is key to the success of managing OKD and FCOS updates together.
The machine-config-server daemon set, which provides the Ignition config files to control plane nodes as they join the cluster.

The machine configuration is a subset of the Ignition configuration. The machine-config-daemon reads the machine configuration to see if it needs to do an OSTree update or if it must apply a series of systemd kubelet file changes, configuration changes, or other changes to the operating system or OKD configuration.

When you perform node management operations, you create or modify a KubeletConfig custom resource (CR).

When changes are made to a machine configuration, the Machine Config Operator (MCO) automatically reboots all corresponding nodes in order for the changes to take effect.

To prevent the nodes from automatically rebooting after machine configuration changes, before making the changes, you must pause the autoreboot process by setting the spec.paused field to true in the corresponding machine config pool. When paused, machine configuration changes are not applied until you set the spec.paused field to false and the nodes have rebooted into the new configuration.

The following modifications do not trigger a node reboot:

When the MCO detects any of the following changes, it applies the update without draining or rebooting the node:
- Changes to the SSH key in the spec.config.passwd.users.sshAuthorizedKeys parameter of a machine config.
- Changes to the global pull secret or pull secret in the openshift-config namespace.
- Automatic rotation of the /etc/kubernetes/kubelet-ca.crt certificate authority (CA) by the Kubernetes API Server Operator.
When the MCO detects changes to the /etc/containers/registries.conf file, such as adding or editing an ImageDigestMirrorSet, ImageTagMirrorSet, or ImageContentSourcePolicy object, it drains the corresponding nodes, applies the changes, and uncordons the nodes. The node drain does not happen for the following changes:
- The addition of a registry with the pull-from-mirror = "digest-only" parameter set for each mirror.
- The addition of a mirror with the pull-from-mirror = "digest-only" parameter set in a registry.
- The addition of items to the unqualified-search-registries list.

There might be situations where the configuration on a node does not fully match what the currently-applied machine config specifies. This state is called configuration drift. The Machine Config Daemon (MCD) regularly checks the nodes for configuration drift. If the MCD detects configuration drift, the MCO marks the node degraded until an administrator corrects the node configuration. A degraded node is online and operational, but, it cannot be updated.

Machine Config overview

The Machine Config Operator (MCO) manages updates to systemd, CRI-O and Kubelet, the kernel, Network Manager and other system features. It also offers a MachineConfig CRD that can write configuration files onto the host (see machine-config-operator). Understanding what the MCO does and how it interacts with other components is critical to making advanced, system-level changes to an OKD cluster. Here are some things you should know about the MCO, machine configs, and how they are used:

Machine configs are processed alphabetically, in lexicographically increasing order, of their name. The render controller uses the first machine config in the list as the base and appends the rest to the base machine config.
A machine config can make a specific change to a file or service on the operating system of each system representing a pool of OKD nodes.
MCO applies changes to operating systems in pools of machines. All OKD clusters start with worker and control plane node pools. By adding more role labels, you can configure custom pools of nodes. For example, you can set up a custom pool of worker nodes that includes particular hardware features needed by an application. However, examples in this section focus on changes to the default pool types.

A node can have multiple labels applied that indicate its type, such as master or worker, however it can be a member of only a single machine config pool.
After a machine config change, the MCO updates the affected nodes alphabetically by zone, based on the topology.kubernetes.io/zone label. If a zone has more than one node, the oldest nodes are updated first. For nodes that do not use zones, such as in bare metal deployments, the nodes are upgraded by age, with the oldest nodes updated first. The MCO updates the number of nodes as specified by the maxUnavailable field on the machine configuration pool at a time.
Some machine configuration must be in place before OKD is installed to disk. In most cases, this can be accomplished by creating a machine config that is injected directly into the OKD installer process, instead of running as a postinstallation machine config. In other cases, you might need to do bare metal installation where you pass kernel arguments at OKD installer startup, to do such things as setting per-node individual IP addresses or advanced disk partitioning.
MCO manages items that are set in machine configs. Manual changes you do to your systems will not be overwritten by MCO, unless MCO is explicitly told to manage a conflicting file. In other words, MCO only makes specific updates you request, it does not claim control over the whole node.
Manual changes to nodes are strongly discouraged. If you need to decommission a node and start a new one, those direct changes would be lost.
MCO is only supported for writing to files in /etc and /var directories, although there are symbolic links to some directories that can be writeable by being symbolically linked to one of those areas. The /opt and /usr/local directories are examples.
Ignition is the configuration format used in MachineConfigs. See the Ignition Configuration Specification v3.2.0 for details.
Although Ignition config settings can be delivered directly at OKD installation time, and are formatted in the same way that MCO delivers Ignition configs, MCO has no way of seeing what those original Ignition configs are. Therefore, you should wrap Ignition config settings into a machine config before deploying them.
When a file managed by MCO changes outside of MCO, the Machine Config Daemon (MCD) sets the node as degraded. It will not overwrite the offending file, however, and should continue to operate in a degraded state.
A key reason for using a machine config is that it will be applied when you spin up new nodes for a pool in your OKD cluster. The machine-api-operator provisions a new machine and MCO configures it.

MCO uses Ignition as the configuration format. OKD 4.6 moved from Ignition config specification version 2 to version 3.

What can you change with machine configs?

The kinds of components that MCO can change include:

config: Create Ignition config objects (see the Ignition configuration specification) to do things like modify files, systemd services, and other features on OKD machines, including:
- Configuration files: Create or overwrite files in the /var or /etc directory.
- systemd units: Create and set the status of a systemd service or add to an existing systemd service by dropping in additional settings.
- users and groups: Change SSH keys in the passwd section postinstallation.
  Changing SSH keys by using a machine config is supported only for the core user.
  
  Adding new users by using a machine config is not supported.
kernelArguments: Add arguments to the kernel command line when OKD nodes boot.
kernelType: Optionally identify a non-standard kernel to use instead of the standard kernel. Use realtime to use the RT kernel (for RAN). This is only supported on select platforms.

extensions: Extend FCOS features by adding selected pre-packaged software. For this feature, available extensions include usbguard and kernel modules.
Custom resources (for ContainerRuntime and Kubelet): Outside of machine configs, MCO manages two special custom resources for modifying CRI-O container runtime settings (ContainerRuntime CR) and the Kubelet service (Kubelet CR).

The MCO is not the only Operator that can change operating system components on OKD nodes. Other Operators can modify operating system-level features as well. One example is the Node Tuning Operator, which allows you to do node-level tuning through Tuned daemon profiles.

Tasks for the MCO configuration that can be done postinstallation are included in the following procedures. See descriptions of FCOS bare metal installation for system configuration tasks that must be done during or before OKD installation.

Project

See the openshift-machine-config-operator GitHub site for details.

Understanding the Machine Config Operator node drain behavior

When you use a machine config to change a system feature, such as adding new config files, modifying systemd units or kernel arguments, or updating SSH keys, the Machine Config Operator (MCO) applies those changes and ensures that each node is in the desired configuration state.

After you make the changes, the MCO generates a new rendered machine config. In the majority of cases, when applying the new rendered machine config, the Operator performs the following steps on each affected node until all of the affected nodes have the updated configuration:

Cordon. The MCO marks the node as not schedulable for additional workloads.
Drain. The MCO terminates all running workloads on the node, causing the workloads to be rescheduled onto other nodes.
Apply. The MCO writes the new configuration to the nodes as needed.
Reboot. The MCO restarts the node.
Uncordon. The MCO marks the node as schedulable for workloads.

Throughout this process, the MCO maintains the required number of pods based on the MaxUnavailable value set in the machine config pool.

If the MCO drains pods on the master node, note the following conditions:

In single-node OpenShift clusters, the MCO skips the drain operation.
The MCO does not drain static pods in order to prevent interference with services, such as etcd.

In certain cases the nodes are not drained. For more information, see "About the Machine Config Operator."

You can mitigate the disruption caused by drain and reboot cycles by disabling control plane reboots. For more information, see "Disabling the Machine Config Operator from automatically rebooting."

Additional resources

Understanding configuration drift detection

There might be situations when the on-disk state of a node differs from what is configured in the machine config. This is known as configuration drift. For example, a cluster admin might manually modify a file, a systemd unit file, or a file permission that was configured through a machine config. This causes configuration drift. Configuration drift can cause problems between nodes in a Machine Config Pool or when the machine configs are updated.

The Machine Config Operator (MCO) uses the Machine Config Daemon (MCD) to check nodes for configuration drift on a regular basis. If detected, the MCO sets the node and the machine config pool (MCP) to Degraded and reports the error. A degraded node is online and operational, but, it cannot be updated.

The MCD performs configuration drift detection upon each of the following conditions:

When a node boots.
After any of the files (Ignition files and systemd drop-in units) specified in the machine config are modified outside of the machine config.

Before a new machine config is applied.

If you apply a new machine config to the nodes, the MCD temporarily shuts down configuration drift detection. This shutdown is needed because the new machine config necessarily differs from the machine config on the nodes. After the new machine config is applied, the MCD restarts detecting configuration drift using the new machine config.

When performing configuration drift detection, the MCD validates that the file contents and permissions fully match what the currently-applied machine config specifies. Typically, the MCD detects configuration drift in less than a second after the detection is triggered.

If the MCD detects configuration drift, the MCD performs the following tasks:

Emits an error to the console logs
Emits a Kubernetes event
Stops further detection on the node
Sets the node and MCP to degraded

You can check if you have a degraded node by listing the MCPs:

$ oc get mcp worker

If you have a degraded MCP, the DEGRADEDMACHINECOUNT field is non-zero, similar to the following output:

Example output

NAME     CONFIG                                             UPDATED   UPDATING   DEGRADED   MACHINECOUNT   READYMACHINECOUNT   UPDATEDMACHINECOUNT   DEGRADEDMACHINECOUNT   AGE
worker   rendered-worker-404caf3180818d8ac1f50c32f14b57c3   False     True       True       2              1                   1                     1                      5h51m

You can determine if the problem is caused by configuration drift by examining the machine config pool:

$ oc describe mcp worker

Example output

 ...
    Last Transition Time:  2021-12-20T18:54:00Z
    Message:               Node ci-ln-j4h8nkb-72292-pxqxz-worker-a-fjks4 is reporting: "content mismatch for file \"/etc/mco-test-file\"" (1)
    Reason:                1 nodes are reporting degraded status on sync
    Status:                True
    Type:                  NodeDegraded (2)
 ...

1	This message shows that a node’s `/etc/mco-test-file` file, which was added by the machine config, has changed outside of the machine config.
2	The state of the node is `NodeDegraded`.

Or, if you know which node is degraded, examine that node:

$ oc describe node/ci-ln-j4h8nkb-72292-pxqxz-worker-a-fjks4

Example output

 ...

Annotations:        cloud.network.openshift.io/egress-ipconfig: [{"interface":"nic0","ifaddr":{"ipv4":"10.0.128.0/17"},"capacity":{"ip":10}}]
                    csi.volume.kubernetes.io/nodeid:
                      {"pd.csi.storage.gke.io":"projects/openshift-gce-devel-ci/zones/us-central1-a/instances/ci-ln-j4h8nkb-72292-pxqxz-worker-a-fjks4"}
                    machine.openshift.io/machine: openshift-machine-api/ci-ln-j4h8nkb-72292-pxqxz-worker-a-fjks4
                    machineconfiguration.openshift.io/controlPlaneTopology: HighlyAvailable
                    machineconfiguration.openshift.io/currentConfig: rendered-worker-67bd55d0b02b0f659aef33680693a9f9
                    machineconfiguration.openshift.io/desiredConfig: rendered-worker-67bd55d0b02b0f659aef33680693a9f9
                    machineconfiguration.openshift.io/reason: content mismatch for file "/etc/mco-test-file" (1)
                    machineconfiguration.openshift.io/state: Degraded (2)
 ...

1	The error message indicating that configuration drift was detected between the node and the listed machine config. Here the error message indicates that the contents of the `/etc/mco-test-file`, which was added by the machine config, has changed outside of the machine config.
2	The state of the node is `Degraded`.

You can correct configuration drift and return the node to the Ready state by performing one of the following remediations:

Ensure that the contents and file permissions of the files on the node match what is configured in the machine config. You can manually rewrite the file contents or change the file permissions.
Generate a force file on the degraded node. The force file causes the MCD to bypass the usual configuration drift detection and reapplies the current machine config.

Generating a force file on a node causes that node to reboot.

Additional resources

Disabling Machine Config Operator from automatically rebooting.

Checking machine config pool status

To see the status of the Machine Config Operator (MCO), its sub-components, and the resources it manages, use the following oc commands:

Procedure

To see the number of MCO-managed nodes available on your cluster for each machine config pool (MCP), run the following command:

$ oc get machineconfigpool

Example output

NAME      CONFIG                    UPDATED  UPDATING   DEGRADED  MACHINECOUNT  READYMACHINECOUNT  UPDATEDMACHINECOUNT DEGRADEDMACHINECOUNT  AGE
master    rendered-master-06c9c4…   True     False      False     3             3                  3                   0                     4h42m
worker    rendered-worker-f4b64…    False    True       False     3             2                  2                   0                     4h42m

where:

UPDATED: The True status indicates that the MCO has applied the current machine config to the nodes in that MCP. The current machine config is specified in the STATUS field in the oc get mcp output. The False status indicates a node in the MCP is updating.
UPDATING: The True status indicates that the MCO is applying the desired machine config, as specified in the MachineConfigPool custom resource, to at least one of the nodes in that MCP. The desired machine config is the new, edited machine config. Nodes that are updating might not be available for scheduling. The False status indicates that all nodes in the MCP are updated.
DEGRADED: A True status indicates the MCO is blocked from applying the current or desired machine config to at least one of the nodes in that MCP, or the configuration is failing. Nodes that are degraded might not be available for scheduling. A False status indicates that all nodes in the MCP are ready.
MACHINECOUNT: Indicates the total number of machines in that MCP.
READYMACHINECOUNT: Indicates the total number of machines in that MCP that are ready for scheduling.
UPDATEDMACHINECOUNT: Indicates the total number of machines in that MCP that have the current machine config.
DEGRADEDMACHINECOUNT: Indicates the total number of machines in that MCP that are marked as degraded or unreconcilable.

In the previous output, there are three control plane (master) nodes and three worker nodes. The control plane MCP and the associated nodes are updated to the current machine config. The nodes in the worker MCP are being updated to the desired machine config. Two of the nodes in the worker MCP are updated and one is still updating, as indicated by the UPDATEDMACHINECOUNT being 2. There are no issues, as indicated by the DEGRADEDMACHINECOUNT being 0 and DEGRADED being False.

While the nodes in the MCP are updating, the machine config listed under CONFIG is the current machine config, which the MCP is being updated from. When the update is complete, the listed machine config is the desired machine config, which the MCP was updated to.

If a node is being cordoned, that node is not included in the READYMACHINECOUNT, but is included in the MACHINECOUNT. Also, the MCP status is set to UPDATING. Because the node has the current machine config, it is counted in the UPDATEDMACHINECOUNT total:

Example output

NAME      CONFIG                    UPDATED  UPDATING   DEGRADED  MACHINECOUNT  READYMACHINECOUNT  UPDATEDMACHINECOUNT DEGRADEDMACHINECOUNT  AGE
master    rendered-master-06c9c4…   True     False      False     3             3                  3                   0                     4h42m
worker    rendered-worker-c1b41a…   False    True       False     3             2                  3                   0                     4h42m

To check the status of the nodes in an MCP by examining the MachineConfigPool custom resource, run the following command: :

$ oc describe mcp worker

Example output

...
  Degraded Machine Count:     0
  Machine Count:              3
  Observed Generation:        2
  Ready Machine Count:        3
  Unavailable Machine Count:  0
  Updated Machine Count:      3
Events:                       <none>

If a node is being cordoned, the node is not included in the Ready Machine Count. It is included in the Unavailable Machine Count:

Example output

...
  Degraded Machine Count:     0
  Machine Count:              3
  Observed Generation:        2
  Ready Machine Count:        2
  Unavailable Machine Count:  1
  Updated Machine Count:      3

To see each existing MachineConfig object, run the following command:

$ oc get machineconfigs

Example output

NAME                             GENERATEDBYCONTROLLER          IGNITIONVERSION  AGE
00-master                        2c9371fbb673b97a6fe8b1c52...   3.2.0            5h18m
00-worker                        2c9371fbb673b97a6fe8b1c52...   3.2.0            5h18m
01-master-container-runtime      2c9371fbb673b97a6fe8b1c52...   3.2.0            5h18m
01-master-kubelet                2c9371fbb673b97a6fe8b1c52…     3.2.0            5h18m
...
rendered-master-dde...           2c9371fbb673b97a6fe8b1c52...   3.2.0            5h18m
rendered-worker-fde...           2c9371fbb673b97a6fe8b1c52...   3.2.0            5h18m

Note that the MachineConfig objects listed as rendered are not meant to be changed or deleted.

To view the contents of a particular machine config (in this case, 01-master-kubelet), run the following command:

$ oc describe machineconfigs 01-master-kubelet

The output from the command shows that this MachineConfig object contains both configuration files (cloud.conf and kubelet.conf) and a systemd service (Kubernetes Kubelet):

Example output

Name:         01-master-kubelet
...
Spec:
  Config:
    Ignition:
      Version:  3.2.0
    Storage:
      Files:
        Contents:
          Source:   data:,
        Mode:       420
        Overwrite:  true
        Path:       /etc/kubernetes/cloud.conf
        Contents:
          Source:   data:,kind%3A%20KubeletConfiguration%0AapiVersion%3A%20kubelet.config.k8s.io%2Fv1beta1%0Aauthentication%3A%0A%20%20x509%3A%0A%20%20%20%20clientCAFile%3A%20%2Fetc%2Fkubernetes%2Fkubelet-ca.crt%0A%20%20anonymous...
        Mode:       420
        Overwrite:  true
        Path:       /etc/kubernetes/kubelet.conf
    Systemd:
      Units:
        Contents:  [Unit]
Description=Kubernetes Kubelet
Wants=rpc-statd.service network-online.target crio.service
After=network-online.target crio.service

ExecStart=/usr/bin/hyperkube \
    kubelet \
      --config=/etc/kubernetes/kubelet.conf \ ...

If something goes wrong with a machine config that you apply, you can always back out that change. For example, if you had run oc create -f ./myconfig.yaml to apply a machine config, you could remove that machine config by running the following command:

$ oc delete -f ./myconfig.yaml

If that was the only problem, the nodes in the affected pool should return to a non-degraded state. This actually causes the rendered configuration to roll back to its previously rendered state.

If you add your own machine configs to your cluster, you can use the commands shown in the previous example to check their status and the related status of the pool to which they are applied.

Using MachineConfig objects to configure nodes

You can use the tasks in this section to create MachineConfig objects that modify files, systemd unit files, and other operating system features running on OKD nodes. For more ideas on working with machine configs, see content related to updating SSH authorized keys, verifying image signatures, enabling SCTP, and configuring iSCSI initiatornames for OKD.

OKD supports Ignition specification version 3.2. All new machine configs you create going forward should be based on Ignition specification version 3.2. If you are upgrading your OKD cluster, any existing Ignition specification version 2.x machine configs will be translated automatically to specification version 3.2.

Use the following "Configuring chrony time service" procedure as a model for how to go about adding other configuration files to OKD nodes.

Configuring chrony time service

You can set the time server and related settings used by the chrony time service (chronyd) by modifying the contents of the chrony.conf file and passing those contents to your nodes as a machine config.

Procedure

Create a Butane config including the contents of the chrony.conf file. For example, to configure chrony on worker nodes, create a 99-worker-chrony.bu file.

See "Creating machine configs with Butane" for information about Butane.

variant: openshift
version: 4.13.0
metadata:
  name: 99-worker-chrony (1)
  labels:
    machineconfiguration.openshift.io/role: worker (1)
storage:
  files:
  - path: /etc/chrony.conf
    mode: 0644 (2)
    overwrite: true
    contents:
      inline: |
        pool 0.rhel.pool.ntp.org iburst (3)
        driftfile /var/lib/chrony/drift
        makestep 1.0 3
        rtcsync
        logdir /var/log/chrony

1	On control plane nodes, substitute `master` for `worker` in both of these locations.
2	Specify an octal value mode for the `mode` field in the machine config file. After creating the file and applying the changes, the `mode` is converted to a decimal value. You can check the YAML file with the command `oc get mc <mc-name> -o yaml`.
3	Specify any valid, reachable time source, such as the one provided by your DHCP server. Alternately, you can specify any of the following NTP servers: `1.rhel.pool.ntp.org`, `2.rhel.pool.ntp.org`, or `3.rhel.pool.ntp.org`.

Use Butane to generate a MachineConfig object file, 99-worker-chrony.yaml, containing the configuration to be delivered to the nodes:
```
$ butane 99-worker-chrony.bu -o 99-worker-chrony.yaml
```
Apply the configurations in one of two ways:
- If the cluster is not running yet, after you generate manifest files, add the MachineConfig object file to the <installation_directory>/openshift directory, and then continue to create the cluster.
- If the cluster is already running, apply the file:
  $ oc apply -f ./99-worker-chrony.yaml

Additional resources

Creating machine configs with Butane

Disabling the chrony time service

You can disable the chrony time service (chronyd) for nodes with a specific role by using a MachineConfig custom resource (CR).

Prerequisites

Install the OpenShift CLI (oc).
Log in as a user with cluster-admin privileges.

Procedure

Create the MachineConfig CR that disables chronyd for the specified node role.

Save the following YAML in the disable-chronyd.yaml file:

apiVersion: machineconfiguration.openshift.io/v1
kind: MachineConfig
metadata:
  labels:
    machineconfiguration.openshift.io/role: <node_role> (1)
  name: disable-chronyd
spec:
  config:
    ignition:
      version: 3.2.0
    systemd:
      units:
        - contents: |
            [Unit]
            Description=NTP client/server
            Documentation=man:chronyd(8) man:chrony.conf(5)
            After=ntpdate.service sntp.service ntpd.service
            Conflicts=ntpd.service systemd-timesyncd.service
            ConditionCapability=CAP_SYS_TIME
            [Service]
            Type=forking
            PIDFile=/run/chrony/chronyd.pid
            EnvironmentFile=-/etc/sysconfig/chronyd
            ExecStart=/usr/sbin/chronyd $OPTIONS
            ExecStartPost=/usr/libexec/chrony-helper update-daemon
            PrivateTmp=yes
            ProtectHome=yes
            ProtectSystem=full
            [Install]
            WantedBy=multi-user.target
          enabled: false
          name: "chronyd.service"

1	Node role where you want to disable `chronyd`, for example, `master`.

Create the MachineConfig CR by running the following command:
```
$ oc create -f disable-chronyd.yaml
```

Adding kernel arguments to nodes

In some special cases, you might want to add kernel arguments to a set of nodes in your cluster. This should only be done with caution and clear understanding of the implications of the arguments you set.

Improper use of kernel arguments can result in your systems becoming unbootable.

Examples of kernel arguments you could set include:

nosmt: Disables symmetric multithreading (SMT) in the kernel. Multithreading allows multiple logical threads for each CPU. You could consider nosmt in multi-tenant environments to reduce risks from potential cross-thread attacks. By disabling SMT, you essentially choose security over performance.
systemd.unified_cgroup_hierarchy: Configures the version of Linux control group that is installed on your nodes: cgroup v1 or cgroup v2. cgroup v2 is the next version of the kernel control group and offers multiple improvements. However, it can have some unwanted effects on your nodes.

cgroup v2 is enabled by default. To disable cgroup v2, use the systemd.unified_cgroup_hierarchy=0 kernel argument, as shown in the following procedure.
enforcing=0: Configures Security Enhanced Linux (SELinux) to run in permissive mode. In permissive mode, the system acts as if SELinux is enforcing the loaded security policy, including labeling objects and emitting access denial entries in the logs, but it does not actually deny any operations. While not supported for production systems, permissive mode can be helpful for debugging.

Disabling SELinux on FCOS in production is not supported. Once SELinux has been disabled on a node, it must be re-provisioned before re-inclusion in a production cluster.

See Kernel.org kernel parameters for a list and descriptions of kernel arguments.

In the following procedure, you create a MachineConfig object that identifies:

A set of machines to which you want to add the kernel argument. In this case, machines with a worker role.
Kernel arguments that are appended to the end of the existing kernel arguments.
A label that indicates where in the list of machine configs the change is applied.

Prerequisites

Have administrative privilege to a working OKD cluster.

Procedure

List existing MachineConfig objects for your OKD cluster to determine how to label your machine config:

$ oc get MachineConfig

Example output

NAME                                               GENERATEDBYCONTROLLER                      IGNITIONVERSION   AGE
00-master                                          52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
00-worker                                          52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
01-master-container-runtime                        52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
01-master-kubelet                                  52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
01-worker-container-runtime                        52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
01-worker-kubelet                                  52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
99-master-generated-registries                     52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
99-master-ssh                                                                                 3.2.0             40m
99-worker-generated-registries                     52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
99-worker-ssh                                                                                 3.2.0             40m
rendered-master-23e785de7587df95a4b517e0647e5ab7   52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
rendered-worker-5d596d9293ca3ea80c896a1191735bb1   52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m

Create a MachineConfig object file that identifies the kernel argument (for example, 05-worker-kernelarg-selinuxpermissive.yaml)

apiVersion: machineconfiguration.openshift.io/v1
kind: MachineConfig
metadata:
  labels:
    machineconfiguration.openshift.io/role: worker (1)
  name: 05-worker-kernelarg-selinuxpermissive (2)
spec:
  config:
    ignition:
      version: 3.2.0
  kernelArguments:
    - enforcing=0 (3)
      systemd.unified_cgroup_hierarchy=0 (4)
#...

1	Applies the new kernel argument only to worker nodes.
2	Named to identify where it fits among the machine configs (05) and what it does (adds a kernel argument to configure SELinux permissive mode).
3	Identifies the exact kernel argument as `enforcing=0`.
4	Configures cgroup v1 on the associated nodes. cgroup v2 is the default.

Create the new machine config:

$ oc create -f 05-worker-kernelarg-selinuxpermissive.yaml

Check the machine configs to see that the new one was added:

$ oc get MachineConfig

Example output

NAME                                               GENERATEDBYCONTROLLER                      IGNITIONVERSION   AGE
00-master                                          52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
00-worker                                          52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
01-master-container-runtime                        52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
01-master-kubelet                                  52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
01-worker-container-runtime                        52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
01-worker-kubelet                                  52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
05-worker-kernelarg-selinuxpermissive                                                         3.2.0             105s
99-master-generated-registries                     52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
99-master-ssh                                                                                 3.2.0             40m
99-worker-generated-registries                     52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
99-worker-ssh                                                                                 3.2.0             40m
rendered-master-23e785de7587df95a4b517e0647e5ab7   52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
rendered-worker-5d596d9293ca3ea80c896a1191735bb1   52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m

Check the nodes:

$ oc get nodes

Example output

NAME                           STATUS                     ROLES    AGE   VERSION
ip-10-0-136-161.ec2.internal   Ready                      worker   28m   v1.26.0
ip-10-0-136-243.ec2.internal   Ready                      master   34m   v1.26.0
ip-10-0-141-105.ec2.internal   Ready,SchedulingDisabled   worker   28m   v1.26.0
ip-10-0-142-249.ec2.internal   Ready                      master   34m   v1.26.0
ip-10-0-153-11.ec2.internal    Ready                      worker   28m   v1.26.0
ip-10-0-153-150.ec2.internal   Ready                      master   34m   v1.26.0

You can see that scheduling on each worker node is disabled as the change is being applied.

Check that the kernel argument worked by going to one of the worker nodes and listing the kernel command line arguments (in /proc/cmdline on the host):

$ oc debug node/ip-10-0-141-105.ec2.internal

Example output

Starting pod/ip-10-0-141-105ec2internal-debug ...
To use host binaries, run `chroot /host`

sh-4.2# cat /host/proc/cmdline
BOOT_IMAGE=/ostree/rhcos-... console=tty0 console=ttyS0,115200n8
rootflags=defaults,prjquota rw root=UUID=fd0... ostree=/ostree/boot.0/rhcos/16...
coreos.oem.id=qemu coreos.oem.id=ec2 ignition.platform.id=ec2 enforcing=0

sh-4.2# exit

You should see the enforcing=0 argument added to the other kernel arguments.

Enabling multipathing with kernel arguments on FCOS

Fedora CoreOS (FCOS) supports multipathing on the primary disk, allowing stronger resilience to hardware failure to achieve higher host availability. Postinstallation support is available by activating multipathing via the machine config.

Enabling multipathing during installation is supported and recommended for nodes provisioned in OKD 4.8 or higher. In setups where any I/O to non-optimized paths results in I/O system errors, you must enable multipathing at installation time. For more information about enabling multipathing during installation time, see "Enabling multipathing with kernel arguments on RHCOS" in the Installing on bare metal documentation.

On IBM Z and IBM® LinuxONE, you can enable multipathing only if you configured your cluster for it during installation. For more information, see "Installing FCOS and starting the OKD bootstrap process" in Installing a cluster with z/VM on IBM Z and IBM® LinuxONE.

When a OKD 4.13 cluster is installed or configured as a post-installation activity on a single VIOS host with "vSCSI" storage on IBM Power® with multipath configured, the CoreOS nodes with multipath enabled fail to boot. This behavior is expected, as only one path is available to the Node.

Prerequisites

You have a running OKD cluster that uses version 4.7 or later.
You are logged in to the cluster as a user with administrative privileges.
You have confirmed that the disk is enabled for multipathing. Multipathing is only supported on hosts that are connected to a SAN via an HBA adapter.

Procedure

To enable multipathing postinstallation on control plane nodes:

Create a machine config file, such as 99-master-kargs-mpath.yaml, that instructs the cluster to add the master label and that identifies the multipath kernel argument, for example:

apiVersion: machineconfiguration.openshift.io/v1
kind: MachineConfig
metadata:
  labels:
    machineconfiguration.openshift.io/role: "master"
  name: 99-master-kargs-mpath
spec:
  kernelArguments:
    - 'rd.multipath=default'
    - 'root=/dev/disk/by-label/dm-mpath-root'

To enable multipathing postinstallation on worker nodes:

Create a machine config file, such as 99-worker-kargs-mpath.yaml, that instructs the cluster to add the worker label and that identifies the multipath kernel argument, for example:

apiVersion: machineconfiguration.openshift.io/v1
kind: MachineConfig
metadata:
  labels:
    machineconfiguration.openshift.io/role: "worker"
  name: 99-worker-kargs-mpath
spec:
  kernelArguments:
    - 'rd.multipath=default'
    - 'root=/dev/disk/by-label/dm-mpath-root'

Create the new machine config by using either the master or worker YAML file you previously created:
```
$ oc create -f ./99-worker-kargs-mpath.yaml
```

Check the machine configs to see that the new one was added:

$ oc get MachineConfig

Example output

NAME                                               GENERATEDBYCONTROLLER                      IGNITIONVERSION   AGE
00-master                                          52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
00-worker                                          52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
01-master-container-runtime                        52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
01-master-kubelet                                  52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
01-worker-container-runtime                        52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
01-worker-kubelet                                  52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
99-master-generated-registries                     52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
99-master-ssh                                                                                 3.2.0             40m
99-worker-generated-registries                     52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
99-worker-kargs-mpath                              52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             105s
99-worker-ssh                                                                                 3.2.0             40m
rendered-master-23e785de7587df95a4b517e0647e5ab7   52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m
rendered-worker-5d596d9293ca3ea80c896a1191735bb1   52dd3ba6a9a527fc3ab42afac8d12b693534c8c9   3.2.0             33m

Check the nodes:

$ oc get nodes

Example output

NAME                           STATUS                     ROLES    AGE   VERSION
ip-10-0-136-161.ec2.internal   Ready                      worker   28m   v1.26.0
ip-10-0-136-243.ec2.internal   Ready                      master   34m   v1.26.0
ip-10-0-141-105.ec2.internal   Ready,SchedulingDisabled   worker   28m   v1.26.0
ip-10-0-142-249.ec2.internal   Ready                      master   34m   v1.26.0
ip-10-0-153-11.ec2.internal    Ready                      worker   28m   v1.26.0
ip-10-0-153-150.ec2.internal   Ready                      master   34m   v1.26.0

You can see that scheduling on each worker node is disabled as the change is being applied.

Check that the kernel argument worked by going to one of the worker nodes and listing the kernel command line arguments (in /proc/cmdline on the host):

$ oc debug node/ip-10-0-141-105.ec2.internal

Example output

Starting pod/ip-10-0-141-105ec2internal-debug ...
To use host binaries, run `chroot /host`

sh-4.2# cat /host/proc/cmdline
...
rd.multipath=default root=/dev/disk/by-label/dm-mpath-root
...

sh-4.2# exit

You should see the added kernel arguments.

Additional resources

See Enabling multipathing with kernel arguments on RHCOS for more information about enabling multipathing during installation time.

Adding a real-time kernel to nodes

Some OKD workloads require a high degree of determinism.While Linux is not a real-time operating system, the Linux real-time kernel includes a preemptive scheduler that provides the operating system with real-time characteristics.

If your OKD workloads require these real-time characteristics, you can switch your machines to the Linux real-time kernel. For OKD, 4.13 you can make this switch using a MachineConfig object. Although making the change is as simple as changing a machine config kernelType setting to realtime, there are a few other considerations before making the change:

Currently, real-time kernel is supported only on worker nodes, and only for radio access network (RAN) use.
The following procedure is fully supported with bare metal installations that use systems that are certified for Red Hat Enterprise Linux for Real Time 8.
Real-time support in OKD is limited to specific subscriptions.
The following procedure is also supported for use with Google Cloud Platform.

Prerequisites

Have a running OKD cluster (version 4.4 or later).
Log in to the cluster as a user with administrative privileges.

Procedure

Create a machine config for the real-time kernel: Create a YAML file (for example, 99-worker-realtime.yaml) that contains a MachineConfig object for the realtime kernel type. This example tells the cluster to use a real-time kernel for all worker nodes:
```
$ cat << EOF > 99-worker-realtime.yaml
apiVersion: machineconfiguration.openshift.io/v1
kind: MachineConfig
metadata:
  labels:
    machineconfiguration.openshift.io/role: "worker"
  name: 99-worker-realtime
spec:
  kernelType: realtime
EOF
```
Add the machine config to the cluster. Type the following to add the machine config to the cluster:
```
$ oc create -f 99-worker-realtime.yaml
```

Check the real-time kernel: Once each impacted node reboots, log in to the cluster and run the following commands to make sure that the real-time kernel has replaced the regular kernel for the set of nodes you configured:

$ oc get nodes

Example output

NAME                                        STATUS  ROLES    AGE   VERSION
ip-10-0-143-147.us-east-2.compute.internal  Ready   worker   103m  v1.26.0
ip-10-0-146-92.us-east-2.compute.internal   Ready   worker   101m  v1.26.0
ip-10-0-169-2.us-east-2.compute.internal    Ready   worker   102m  v1.26.0

$ oc debug node/ip-10-0-143-147.us-east-2.compute.internal

Example output

Starting pod/ip-10-0-143-147us-east-2computeinternal-debug ...
To use host binaries, run `chroot /host`

sh-4.4# uname -a
Linux <worker_node> 4.18.0-147.3.1.rt24.96.el8_1.x86_64 #1 SMP PREEMPT RT
        Wed Nov 27 18:29:55 UTC 2019 x86_64 x86_64 x86_64 GNU/Linux

The kernel name contains rt and text “PREEMPT RT” indicates that this is a real-time kernel.

To go back to the regular kernel, delete the MachineConfig object:
```
$ oc delete -f 99-worker-realtime.yaml
```

Configuring journald settings

If you need to configure settings for the journald service on OKD nodes, you can do that by modifying the appropriate configuration file and passing the file to the appropriate pool of nodes as a machine config.

This procedure describes how to modify journald rate limiting settings in the /etc/systemd/journald.conf file and apply them to worker nodes. See the journald.conf man page for information on how to use that file.

Prerequisites

Have a running OKD cluster.
Log in to the cluster as a user with administrative privileges.

Procedure

Create a Butane config file, 40-worker-custom-journald.bu, that includes an /etc/systemd/journald.conf file with the required settings.

See "Creating machine configs with Butane" for information about Butane.

variant: openshift
version: 4.13.0
metadata:
  name: 40-worker-custom-journald
  labels:
    machineconfiguration.openshift.io/role: worker
storage:
  files:
  - path: /etc/systemd/journald.conf
    mode: 0644
    overwrite: true
    contents:
      inline: |
        # Disable rate limiting
        RateLimitInterval=1s
        RateLimitBurst=10000
        Storage=volatile
        Compress=no
        MaxRetentionSec=30s

Use Butane to generate a MachineConfig object file, 40-worker-custom-journald.yaml, containing the configuration to be delivered to the worker nodes:
```
$ butane 40-worker-custom-journald.bu -o 40-worker-custom-journald.yaml
```

Apply the machine config to the pool:

$ oc apply -f 40-worker-custom-journald.yaml

Check that the new machine config is applied and that the nodes are not in a degraded state. It might take a few minutes. The worker pool will show the updates in progress, as each node successfully has the new machine config applied:

$ oc get machineconfigpool
NAME   CONFIG             UPDATED UPDATING DEGRADED MACHINECOUNT READYMACHINECOUNT UPDATEDMACHINECOUNT DEGRADEDMACHINECOUNT AGE
master rendered-master-35 True    False    False    3            3                 3                   0                    34m
worker rendered-worker-d8 False   True     False    3            1                 1                   0                    34m

To check that the change was applied, you can log in to a worker node:

$ oc get node | grep worker
ip-10-0-0-1.us-east-2.compute.internal   Ready    worker   39m   v0.0.0-master+$Format:%h$
$ oc debug node/ip-10-0-0-1.us-east-2.compute.internal
Starting pod/ip-10-0-141-142us-east-2computeinternal-debug ...
...
sh-4.2# chroot /host
sh-4.4# cat /etc/systemd/journald.conf
# Disable rate limiting
RateLimitInterval=1s
RateLimitBurst=10000
Storage=volatile
Compress=no
MaxRetentionSec=30s
sh-4.4# exit

Additional resources

Creating machine configs with Butane

Adding extensions to FCOS

FCOS is a minimal container-oriented RHEL operating system, designed to provide a common set of capabilities to OKD clusters across all platforms. While adding software packages to FCOS systems is generally discouraged, the MCO provides an extensions feature you can use to add a minimal set of features to FCOS nodes.

Currently, the following extensions are available:

usbguard: Adding the usbguard extension protects FCOS systems from attacks from intrusive USB devices. See USBGuard for details.
kerberos: Adding the kerberos extension provides a mechanism that allows both users and machines to identify themselves to the network to receive defined, limited access to the areas and services that an administrator has configured. See Using Kerberos for details, including how to set up a Kerberos client and mount a Kerberized NFS share.

The following procedure describes how to use a machine config to add one or more extensions to your FCOS nodes.

Prerequisites

Have a running OKD cluster (version 4.6 or later).
Log in to the cluster as a user with administrative privileges.

Procedure

Create a machine config for extensions: Create a YAML file (for example, 80-extensions.yaml) that contains a MachineConfig extensions object. This example tells the cluster to add the usbguard extension.

$ cat << EOF > 80-extensions.yaml
apiVersion: machineconfiguration.openshift.io/v1
kind: MachineConfig
metadata:
  labels:
    machineconfiguration.openshift.io/role: worker
  name: 80-worker-extensions
spec:
  config:
    ignition:
      version: 3.2.0
  extensions:
    - usbguard
EOF

Add the machine config to the cluster. Type the following to add the machine config to the cluster:
```
$ oc create -f 80-extensions.yaml
```
This sets all worker nodes to have rpm packages for usbguard installed.

Check that the extensions were applied:

$ oc get machineconfig 80-worker-extensions

Example output

NAME                 GENERATEDBYCONTROLLER IGNITIONVERSION AGE
80-worker-extensions                       3.2.0           57s

Check that the new machine config is now applied and that the nodes are not in a degraded state. It may take a few minutes. The worker pool will show the updates in progress, as each machine successfully has the new machine config applied:

$ oc get machineconfigpool

Example output

NAME   CONFIG             UPDATED UPDATING DEGRADED MACHINECOUNT READYMACHINECOUNT UPDATEDMACHINECOUNT DEGRADEDMACHINECOUNT AGE
master rendered-master-35 True    False    False    3            3                 3                   0                    34m
worker rendered-worker-d8 False   True     False    3            1                 1                   0                    34m

Check the extensions. To check that the extension was applied, run:

$ oc get node | grep worker

Example output

NAME                                        STATUS  ROLES    AGE   VERSION
ip-10-0-169-2.us-east-2.compute.internal    Ready   worker   102m  v1.26.0

$ oc debug node/ip-10-0-169-2.us-east-2.compute.internal

Example output

...
To use host binaries, run `chroot /host`
sh-4.4# chroot /host
sh-4.4# rpm -q usbguard
usbguard-0.7.4-4.el8.x86_64.rpm

Loading custom firmware blobs in the machine config manifest

Because the default location for firmware blobs in /usr/lib is read-only, you can locate a custom firmware blob by updating the search path. This enables you to load local firmware blobs in the machine config manifest when the blobs are not managed by FCOS.

Procedure

Create a Butane config file, 98-worker-firmware-blob.bu, that updates the search path so that it is root-owned and writable to local storage. The following example places the custom blob file from your local workstation onto nodes under /var/lib/firmware.

See "Creating machine configs with Butane" for information about Butane.

Butane config file for custom firmware blob

variant: openshift
version: 4.13.0
metadata:
  labels:
    machineconfiguration.openshift.io/role: worker
  name: 98-worker-firmware-blob
storage:
  files:
  - path: /var/lib/firmware/<package_name> (1)
    contents:
      local: <package_name> (2)
    mode: 0644 (3)
openshift:
  kernel_arguments:
    - 'firmware_class.path=/var/lib/firmware' (4)

1	Sets the path on the node where the firmware package is copied to.
2	Specifies a file with contents that are read from a local file directory on the system running Butane. The path of the local file is relative to a `files-dir` directory, which must be specified by using the `--files-dir` option with Butane in the following step.
3	Sets the permissions for the file on the FCOS node. It is recommended to set `0644` permissions.
4	The `firmware_class.path` parameter customizes the kernel search path of where to look for the custom firmware blob that was copied from your local workstation onto the root file system of the node. This example uses `/var/lib/firmware` as the customized path.

Run Butane to generate a MachineConfig object file that uses a copy of the firmware blob on your local workstation named 98-worker-firmware-blob.yaml. The firmware blob contains the configuration to be delivered to the nodes. The following example uses the --files-dir option to specify the directory on your workstation where the local file or files are located:
```
$ butane 98-worker-firmware-blob.bu -o 98-worker-firmware-blob.yaml --files-dir <directory_including_package_name>
```
Apply the configurations to the nodes in one of two ways:
- If the cluster is not running yet, after you generate manifest files, add the MachineConfig object file to the <installation_directory>/openshift directory, and then continue to create the cluster.
- If the cluster is already running, apply the file:
  $ oc apply -f 98-worker-firmware-blob.yaml
  A MachineConfig object YAML file is created for you to finish configuring your machines.
Save the Butane config in case you need to update the MachineConfig object in the future.

Additional resources

Creating machine configs with Butane

Configuring MCO-related custom resources

Besides managing MachineConfig objects, the MCO manages two custom resources (CRs): KubeletConfig and ContainerRuntimeConfig. Those CRs let you change node-level settings impacting how the Kubelet and CRI-O container runtime services behave.

Creating a KubeletConfig CRD to edit kubelet parameters

The kubelet configuration is currently serialized as an Ignition configuration, so it can be directly edited. However, there is also a new kubelet-config-controller added to the Machine Config Controller (MCC). This lets you use a KubeletConfig custom resource (CR) to edit the kubelet parameters.

As the fields in the kubeletConfig object are passed directly to the kubelet from upstream Kubernetes, the kubelet validates those values directly. Invalid values in the kubeletConfig object might cause cluster nodes to become unavailable. For valid values, see the Kubernetes documentation.

Consider the following guidance:

Edit an existing KubeletConfig CR to modify existing settings or add new settings, instead of creating a CR for each change. It is recommended that you create a CR only to modify a different machine config pool, or for changes that are intended to be temporary, so that you can revert the changes.
Create one KubeletConfig CR for each machine config pool with all the config changes you want for that pool.
As needed, create multiple KubeletConfig CRs with a limit of 10 per cluster. For the first KubeletConfig CR, the Machine Config Operator (MCO) creates a machine config appended with kubelet. With each subsequent CR, the controller creates another kubelet machine config with a numeric suffix. For example, if you have a kubelet machine config with a -2 suffix, the next kubelet machine config is appended with -3.

If you are applying a kubelet or container runtime config to a custom machine config pool, the custom role in the machineConfigSelector must match the name of the custom machine config pool.

For example, because the following custom machine config pool is named infra, the custom role must also be infra:

apiVersion: machineconfiguration.openshift.io/v1
kind: MachineConfigPool
metadata:
  name: infra
spec:
  machineConfigSelector:
    matchExpressions:
      - {key: machineconfiguration.openshift.io/role, operator: In, values: [worker,infra]}
# ...

If you want to delete the machine configs, delete them in reverse order to avoid exceeding the limit. For example, you delete the kubelet-3 machine config before deleting the kubelet-2 machine config.

If you have a machine config with a kubelet-9 suffix, and you create another KubeletConfig CR, a new machine config is not created, even if there are fewer than 10 kubelet machine configs.

Example KubeletConfig CR

$ oc get kubeletconfig

NAME                AGE
set-max-pods        15m

Example showing a KubeletConfig machine config

$ oc get mc | grep kubelet

...
99-worker-generated-kubelet-1                  b5c5119de007945b6fe6fb215db3b8e2ceb12511   3.2.0             26m
...

The following procedure is an example to show how to configure the maximum number of pods per node on the worker nodes.

Prerequisites

Obtain the label associated with the static MachineConfigPool CR for the type of node you want to configure. Perform one of the following steps:

View the machine config pool:

$ oc describe machineconfigpool <name>

For example:

$ oc describe machineconfigpool worker

Example output

apiVersion: machineconfiguration.openshift.io/v1
kind: MachineConfigPool
metadata:
  creationTimestamp: 2019-02-08T14:52:39Z
  generation: 1
  labels:
    custom-kubelet: set-max-pods (1)

1	If a label has been added it appears under `labels`.

If the label is not present, add a key/value pair:

$ oc label machineconfigpool worker custom-kubelet=set-max-pods

Procedure

View the available machine configuration objects that you can select:
```
$ oc get machineconfig
```
By default, the two kubelet-related configs are 01-master-kubelet and 01-worker-kubelet.

Check the current value for the maximum pods per node:

$ oc describe node <node_name>

For example:

$ oc describe node ci-ln-5grqprb-f76d1-ncnqq-worker-a-mdv94

Look for value: pods: <value> in the Allocatable stanza:

Example output

Allocatable:
 attachable-volumes-aws-ebs:  25
 cpu:                         3500m
 hugepages-1Gi:               0
 hugepages-2Mi:               0
 memory:                      15341844Ki
 pods:                        250

Set the maximum pods per node on the worker nodes by creating a custom resource file that contains the kubelet configuration:

Kubelet configurations that target a specific machine config pool also affect any dependent pools. For example, creating a kubelet configuration for the pool containing worker nodes will also apply to any subset pools, including the pool containing infrastructure nodes. To avoid this, you must create a new machine config pool with a selection expression that only includes worker nodes, and have your kubelet configuration target this new pool.

apiVersion: machineconfiguration.openshift.io/v1
kind: KubeletConfig
metadata:
  name: set-max-pods
spec:
  machineConfigPoolSelector:
    matchLabels:
      custom-kubelet: set-max-pods (1)
  kubeletConfig:
    maxPods: 500 (2)

1	Enter the label from the machine config pool.
2	Add the kubelet configuration. In this example, use `maxPods` to set the maximum pods per node.

The rate at which the kubelet talks to the API server depends on queries per second (QPS) and burst values. The default values, 50 for kubeAPIQPS and 100 for kubeAPIBurst, are sufficient if there are limited pods running on each node. It is recommended to update the kubelet QPS and burst rates if there are enough CPU and memory resources on the node.

apiVersion: machineconfiguration.openshift.io/v1
kind: KubeletConfig
metadata:
  name: set-max-pods
spec:
  machineConfigPoolSelector:
    matchLabels:
      custom-kubelet: set-max-pods
  kubeletConfig:
    maxPods: <pod_count>
    kubeAPIBurst: <burst_rate>
    kubeAPIQPS: <QPS>

Update the machine config pool for workers with the label:

$ oc label machineconfigpool worker custom-kubelet=set-max-pods

Create the KubeletConfig object:
```
$ oc create -f change-maxPods-cr.yaml
```
Verify that the KubeletConfig object is created:
```
$ oc get kubeletconfig
```
Example output
```
NAME                AGE
set-max-pods        15m
```
Depending on the number of worker nodes in the cluster, wait for the worker nodes to be rebooted one by one. For a cluster with 3 worker nodes, this could take about 10 to 15 minutes.

Verify that the changes are applied to the node:

Check on a worker node that the maxPods value changed:
```
$ oc describe node <node_name>
```

Locate the Allocatable stanza:

 ...
Allocatable:
  attachable-volumes-gce-pd:  127
  cpu:                        3500m
  ephemeral-storage:          123201474766
  hugepages-1Gi:              0
  hugepages-2Mi:              0
  memory:                     14225400Ki
  pods:                       500 (1)
 ...

1	In this example, the `pods` parameter should report the value you set in the `KubeletConfig` object.

Verify the change in the KubeletConfig object:

$ oc get kubeletconfigs set-max-pods -o yaml

This should show a status of True and type:Success, as shown in the following example:

spec:
  kubeletConfig:
    maxPods: 500
  machineConfigPoolSelector:
    matchLabels:
      custom-kubelet: set-max-pods
status:
  conditions:
  - lastTransitionTime: "2021-06-30T17:04:07Z"
    message: Success
    status: "True"
    type: Success

Creating a ContainerRuntimeConfig CR to edit CRI-O parameters

You can change some of the settings associated with the OKD CRI-O runtime for the nodes associated with a specific machine config pool (MCP). Using a ContainerRuntimeConfig custom resource (CR), you set the configuration values and add a label to match the MCP. The MCO then rebuilds the crio.conf and storage.conf configuration files on the associated nodes with the updated values.

To revert the changes implemented by using a ContainerRuntimeConfig CR, you must delete the CR. Removing the label from the machine config pool does not revert the changes.

You can modify the following settings by using a ContainerRuntimeConfig CR:

PIDs limit: Setting the PIDs limit in the ContainerRuntimeConfig is expected to be deprecated. If PIDs limits are required, it is recommended to use the podPidsLimit field in the KubeletConfig CR instead. The default podPidsLimit value is 4096 and the default pids_limit value is 0. If podPidsLimit is lower than pids_limit then the effective container PIDs limit is defined by the value set in podPidsLimit.

The CRI-O flag is applied on the cgroup of the container, while the Kubelet flag is set on the cgroup of the pod. Please adjust the PIDs limit accordingly.
Log level: The logLevel parameter sets the CRI-O log_level parameter, which is the level of verbosity for log messages. The default is info (log_level = info). Other options include fatal, panic, error, warn, debug, and trace.
Overlay size: The overlaySize parameter sets the CRI-O Overlay storage driver size parameter, which is the maximum size of a container image.
Maximum log size: Setting the maximum log size in the ContainerRuntimeConfig is expected to be deprecated. If a maximum log size is required, it is recommended to use the containerLogMaxSize field in the KubeletConfig CR instead.
Container runtime: The defaultRuntime parameter sets the container runtime to either runc or crun. The default is runc.

You should have one ContainerRuntimeConfig CR for each machine config pool with all the config changes you want for that pool. If you are applying the same content to all the pools, you only need one ContainerRuntimeConfig CR for all the pools.

You should edit an existing ContainerRuntimeConfig CR to modify existing settings or add new settings instead of creating a new CR for each change. It is recommended to create a new ContainerRuntimeConfig CR only to modify a different machine config pool, or for changes that are intended to be temporary so that you can revert the changes.

You can create multiple ContainerRuntimeConfig CRs, as needed, with a limit of 10 per cluster. For the first ContainerRuntimeConfig CR, the MCO creates a machine config appended with containerruntime. With each subsequent CR, the controller creates a new containerruntime machine config with a numeric suffix. For example, if you have a containerruntime machine config with a -2 suffix, the next containerruntime machine config is appended with -3.

If you want to delete the machine configs, you should delete them in reverse order to avoid exceeding the limit. For example, you should delete the containerruntime-3 machine config before deleting the containerruntime-2 machine config.

If you have a machine config with a containerruntime-9 suffix, and you create another ContainerRuntimeConfig CR, a new machine config is not created, even if there are fewer than 10 containerruntime machine configs.

Example showing multiple ContainerRuntimeConfig CRs

$ oc get ctrcfg

Example output

NAME         AGE
ctr-overlay  15m
ctr-level    5m45s

Example showing multiple containerruntime machine configs

$ oc get mc | grep container

Example output

...
01-master-container-runtime                        b5c5119de007945b6fe6fb215db3b8e2ceb12511   3.2.0             57m
...
01-worker-container-runtime                        b5c5119de007945b6fe6fb215db3b8e2ceb12511   3.2.0             57m
...
99-worker-generated-containerruntime               b5c5119de007945b6fe6fb215db3b8e2ceb12511   3.2.0             26m
99-worker-generated-containerruntime-1             b5c5119de007945b6fe6fb215db3b8e2ceb12511   3.2.0             17m
99-worker-generated-containerruntime-2             b5c5119de007945b6fe6fb215db3b8e2ceb12511   3.2.0             7m26s
...

The following example sets the log_level field to debug and sets the overlay size to 8 GB:

Example ContainerRuntimeConfig CR

apiVersion: machineconfiguration.openshift.io/v1
kind: ContainerRuntimeConfig
metadata:
 name: overlay-size
spec:
 machineConfigPoolSelector:
   matchLabels:
     pools.operator.machineconfiguration.openshift.io/worker: '' (1)
 containerRuntimeConfig:
   logLevel: debug (2)
   overlaySize: 8G (3)
   defaultRuntime: "crun" (4)

1	Specifies the machine config pool label. For a container runtime config, the role must match the name of the associated machine config pool.
2	Optional: Specifies the level of verbosity for log messages.
3	Optional: Specifies the maximum size of a container image.
4	Optional: Specifies the container runtime to deploy to new containers. The default value is `runc`.

Procedure

To change CRI-O settings using the ContainerRuntimeConfig CR:

Create a YAML file for the ContainerRuntimeConfig CR:

apiVersion: machineconfiguration.openshift.io/v1
kind: ContainerRuntimeConfig
metadata:
 name: overlay-size
spec:
 machineConfigPoolSelector:
   matchLabels:
     pools.operator.machineconfiguration.openshift.io/worker: '' (1)
 containerRuntimeConfig: (2)
   logLevel: debug
   overlaySize: 8G

1	Specify a label for the machine config pool that you want you want to modify.
2	Set the parameters as needed.

Create the ContainerRuntimeConfig CR:
```
$ oc create -f <file_name>.yaml
```

Verify that the CR is created:

$ oc get ContainerRuntimeConfig

Example output

NAME           AGE
overlay-size   3m19s

Check that a new containerruntime machine config is created:

$ oc get machineconfigs | grep containerrun

Example output

99-worker-generated-containerruntime   2c9371fbb673b97a6fe8b1c52691999ed3a1bfc2  3.2.0  31s

Monitor the machine config pool until all are shown as ready:

$ oc get mcp worker

Example output

NAME    CONFIG               UPDATED  UPDATING  DEGRADED  MACHINECOUNT  READYMACHINECOUNT  UPDATEDMACHINECOUNT  DEGRADEDMACHINECOUNT  AGE
worker  rendered-worker-169  False    True      False     3             1                  1                    0                     9h

Verify that the settings were applied in CRI-O:

Open an oc debug session to a node in the machine config pool and run chroot /host.
```
$ oc debug node/<node_name>
```
```
sh-4.4# chroot /host
```

Verify the changes in the crio.conf file:

sh-4.4# crio config | grep 'log_level'

Example output

log_level = "debug"

Verify the changes in the `storage.conf`file:

sh-4.4# head -n 7 /etc/containers/storage.conf

Example output

[storage]
  driver = "overlay"
  runroot = "/var/run/containers/storage"
  graphroot = "/var/lib/containers/storage"
  [storage.options]
    additionalimagestores = []
    size = "8G"

Setting the default maximum container root partition size for Overlay with CRI-O

The root partition of each container shows all of the available disk space of the underlying host. Follow this guidance to set a maximum partition size for the root disk of all containers.

To configure the maximum Overlay size, as well as other CRI-O options like the log level, you can create the following ContainerRuntimeConfig custom resource definition (CRD):

apiVersion: machineconfiguration.openshift.io/v1
kind: ContainerRuntimeConfig
metadata:
 name: overlay-size
spec:
 machineConfigPoolSelector:
   matchLabels:
     custom-crio: overlay-size
 containerRuntimeConfig:
   logLevel: debug
   overlaySize: 8G

Procedure

Create the configuration object:
```
$ oc apply -f overlaysize.yml
```
To apply the new CRI-O configuration to your worker nodes, edit the worker machine config pool:
```
$ oc edit machineconfigpool worker
```

Add the custom-crio label based on the matchLabels name you set in the ContainerRuntimeConfig CRD:

apiVersion: machineconfiguration.openshift.io/v1
kind: MachineConfigPool
metadata:
  creationTimestamp: "2020-07-09T15:46:34Z"
  generation: 3
  labels:
    custom-crio: overlay-size
    machineconfiguration.openshift.io/mco-built-in: ""

Save the changes, then view the machine configs:

$ oc get machineconfigs

New 99-worker-generated-containerruntime and rendered-worker-xyz objects are created:

Example output

99-worker-generated-containerruntime  4173030d89fbf4a7a0976d1665491a4d9a6e54f1   3.2.0             7m42s
rendered-worker-xyz                   4173030d89fbf4a7a0976d1665491a4d9a6e54f1   3.2.0             7m36s

After those objects are created, monitor the machine config pool for the changes to be applied:

$ oc get mcp worker

The worker nodes show UPDATING as True, as well as the number of machines, the number updated, and other details:

Example output

NAME   CONFIG              UPDATED   UPDATING   DEGRADED  MACHINECOUNT  READYMACHINECOUNT  UPDATEDMACHINECOUNT   DEGRADEDMACHINECOUNT   AGE
worker rendered-worker-xyz False True False     3             2                   2                    0                      20h

When complete, the worker nodes transition back to UPDATING as False, and the UPDATEDMACHINECOUNT number matches the MACHINECOUNT:

Example output

NAME   CONFIG              UPDATED   UPDATING   DEGRADED  MACHINECOUNT  READYMACHINECOUNT  UPDATEDMACHINECOUNT   DEGRADEDMACHINECOUNT   AGE
worker   rendered-worker-xyz   True      False      False      3         3            3             0           20h

Looking at a worker machine, you see that the new 8 GB max size configuration is applied to all of the workers:

Example output

head -n 7 /etc/containers/storage.conf
[storage]
  driver = "overlay"
  runroot = "/var/run/containers/storage"
  graphroot = "/var/lib/containers/storage"
  [storage.options]
    additionalimagestores = []
    size = "8G"

Looking inside a container, you see that the root partition is now 8 GB:

Example output

~ $ df -h
Filesystem                Size      Used Available Use% Mounted on
overlay                   8.0G      8.0K      8.0G   0% /

Postinstallation machine configuration tasks

About the Machine Config Operator

Machine Config overview

What can you change with machine configs?

Project

Understanding the Machine Config Operator node drain behavior

Understanding configuration drift detection

Checking machine config pool status

Using MachineConfig objects to configure nodes

Configuring chrony time service

Disabling the chrony time service

Adding kernel arguments to nodes

Enabling multipathing with kernel arguments on FCOS

Adding a real-time kernel to nodes

Configuring journald settings

Adding extensions to FCOS

Loading custom firmware blobs in the machine config manifest

Configuring MCO-related custom resources

Creating a KubeletConfig CRD to edit kubelet parameters

Creating a ContainerRuntimeConfig CR to edit CRI-O parameters

Setting the default maximum container root partition size for Overlay with CRI-O