Last modified March 6, 2022

Forcing nodes to use legacy control croups (cgroups) v1

Platform support
AWS
  • Beta in v17.0.0
Azure
  • Beta in v17.0.0

Flatcar version 3033.2.0 and above uses control groups (cgroups) v2 by default, which means all nodes will be using cgroups v2 for all containers in Kubernetes, as opposed to cgroups v1 used on previous versions.

To ensure a smooth transition, in case you need time to modify applications to make them compatible with cgroups v2, we provide a mechanism that will allow using cgroups v1 on specific node pools.

Configure node pool with cgroup v1

AWS

To enable cgroups v1, you have to edit the MachineDeployment resource of your cluster using the Management API.

Make sure the resource has the node.giantswarm.io/cgroupv1 annotation. The value can be anything you like, as only the presence of that annotation is checked. Here is an example:

apiVersion: cluster.x-k8s.io/v1alpha3
kind: MachineDeployment
metadata:
  annotations:
    giantswarm.io/docs: https://docs.giantswarm.io/ui-api/management-api/crd/machinedeployments.cluster.x-k8s.io/
    node.giantswarm.io/cgroupv1: ""
  labels:
    cluster.x-k8s.io/cluster-name: abcd1
    giantswarm.io/cluster: abcd1
    giantswarm.io/machine-deployment: u6gw3
    giantswarm.io/organization: giantswarm
    release.giantswarm.io/version: 17.0.0
  name: u6gw3
  namespace: org-giantswarm
spec:
  ...

Alternatively you can use kubectl command to annotate the CR like shown below:

kubectl annotate machinedeployments.cluster.x-k8s.io u6gw3 node.giantswarm.io/cgroupv1=""

In order to apply the changes, rolling of the nodes in the modified node pool is required. Rolling of the nodes can be triggered either by an update or by manually by terminating each node.

We recommend to set the proper annotation on the node pool before upgrading to release v17.0.0 or later.

If you want to disable the feature you must remove the annotation from the MachineDeployment resource.

Azure

To enable cgroups v1, you have to edit the MachinePool resource of your cluster using the Management API.

Make sure the resource has the node.giantswarm.io/cgroupv1 annotation. The value can be anything you like, as only the presence of that annotation is checked. Here is an example:

apiVersion: exp.cluster.x-k8s.io/v1alpha3
kind: MachinePool
metadata:
  annotations:
    cluster.k8s.io/cluster-api-autoscaler-node-group-max-size: "2"
    cluster.k8s.io/cluster-api-autoscaler-node-group-min-size: "2"
    machine-pool.giantswarm.io/name: vaclav-cluster-xacl9-2
    node.giantswarm.io/cgroupv1: ""
  creationTimestamp: null
  labels:
    cluster.x-k8s.io/cluster-name: xacl9
    giantswarm.io/cluster: xacl9
    giantswarm.io/machine-pool: f8ak0
    giantswarm.io/organization: giantswarm
  name: f8ak0
  namespace: org-giantswarm
spec:
  ...

Or you can use the kubectl command to annotate the CR.

kubectl annotate machinepools.exp.cluster.x-k8s.io f8ak0 node.giantswarm.io/cgroupv1=""

In order to apply the changes, rolling of the nodes in the modified node pool is required. Rolling of the nodes can be triggered either by an update or by manually by terminating each node.

We recommend to set the proper annotation on the node pool before updating to the release 17.0.0 or later.

If you want to disable the feature you must remove the annotation from the MachinePool custom resource.