Flatcar Linux Update Operator

Flatcar Linux Update Operator is a node reboot controller for Kubernetes running Flatcar Container Linux images. When a reboot is needed after updating the system via update_engine, the operator will drain the node before rebooting it.

Flatcar Linux Update Operator fulfills the same purpose as locksmith, but has better integration with Kubernetes by explicitly marking a node as unschedulable and deleting pods on the node before rebooting.

Design

Flatcar Linux Update Operator is divided into two parts: update-operator and update-agent.

update-agent runs as a DaemonSet on each node, waiting for a UPDATE_STATUS_UPDATED_NEED_REBOOT signal via D-Bus from update_engine. It will indicate via node annotations that it needs a reboot.

update-operator runs as a Deployment, watching changes to node annotations and reboots the nodes as needed. It coordinates the reboots of multiple nodes in the cluster, ensuring that not too many are rebooting at once.

Currently, update-operator only reboots one node at a time.

Requirements

• A Kubernetes cluster (>= 1.6) running on Flatcar Container Linux
• The update-engine.service systemd unit on each machine should be unmasked, enabled and started in systemd
• The locksmithd.service systemd unit on each machine should be masked and stopped in systemd

To unmask a service, run systemctl unmask <name>. To enable a service, run systemctl enable <name>. To start/stop a service, run systemctl start <name> or systemctl stop <name> respectively.

or using a Container Linux Config file:

systemd:
  units:
    - name: locksmithd.service
      mask: true
    - name: update-engine.service
      enabled: true

Usage

Create the update-operator deployment and update-agent daemonset.

kubectl apply -f examples/deploy -R

Test

To test that it is working, you can SSH to a node and trigger an update check by running update_engine_client -check_for_update or simulate a reboot is needed by running locksmithctl send-need-reboot.

You can also annotate one of your nodes using the command below. Shortly after, you should see the Node being drained as a preparation for the reboot.

export NODE="<node name>"
kubectl annotate node $NODE --overwrite \
    flatcar-linux-update.v1.flatcar-linux.net/reboot-needed="true"