sylr / kured

Kubernetes Reboot Daemon

• Introduction
• Kubernetes & OS Compatibility
• Installation
• Configuration
  • Reboot Sentinel File & Period
  • Blocking Reboots via Alerts
  • Blocking Reboots via Pods
  • Prometheus Metrics
  • Slack Notifications
  • Overriding Lock Configuration
• Operation
  • Testing
  • Disabling Reboots
  • Manual Unlock
• Building
• Frequently Asked/Anticipated Questions
• Getting Help

Introduction

Kured (KUbernetes REboot Daemon) is a Kubernetes daemonset that performs safe automatic node reboots when the need to do so is indicated by the package management system of the underlying OS.

• Watches for the presence of a reboot sentinel e.g. /var/run/reboot-required
• Utilises a lock in the API server to ensure only one node reboots at a time
• Optionally defers reboots in the presence of active Prometheus alerts or selected pods
• Cordons & drains worker nodes before reboot, uncordoning them after

Kubernetes & OS Compatibility

The daemon image contains versions of k8s.io/client-go and the kubectl binary for the purposes of maintaining the lock and draining worker nodes. Kubernetes aims to provide forwards & backwards compatibility of one minor version between client and server:

kured	kubectl	k8s.io/client-go	k8s.io/apimachinery	expected kubernetes compatibility
master	1.14.1	v11.0.0	release-1.14	1.13.x, 1.14.x, 1.15.x
1.2.0	1.13.6	v10.0.0	release-1.13	1.12.x, 1.13.x, 1.14.x
1.1.0	1.12.1	v9.0.0	release-1.12	1.11.x, 1.12.x, 1.13.x
1.0.0	1.7.6	v4.0.0	release-1.7	1.6.x, 1.7.x, 1.8.x

See the release notes for specific version compatibility information, including which combination have been formally tested.

Versions >=1.1.0 enter the host mount namespace to invoke systemctl reboot, so should work on any systemd distribution.

Installation

To obtain a default installation without Prometheus alerting interlock or Slack notifications:

kubectl apply -f https://github.com/weaveworks/kured/releases/download/1.2.0/kured-1.2.0-dockerhub.yaml

If you want to customise the installation, download the manifest and edit it in accordance with the following section before application.

Configuration

The following arguments can be passed to kured via the daemonset pod template:

Flags:
      --alert-filter-regexp regexp.Regexp   alert names to ignore when checking for active alerts
      --blocking-pod-selector stringArray   label selector identifying pods whose presence should prevent reboots
      --ds-name string                      name of daemonset on which to place lock (default "kured")
      --ds-namespace string                 namespace containing daemonset on which to place lock (default "kube-system")
  -h, --help                                help for kured
      --lock-annotation string              annotation in which to record locking node (default "weave.works/kured-node-lock")
      --period duration                     reboot check period (default 1h0m0s)
      --prometheus-url string               Prometheus instance to probe for active alerts
      --reboot-sentinel string              path to file whose existence signals need to reboot (default "/var/run/reboot-required")
      --slack-hook-url string               slack hook URL for reboot notfications
      --slack-username string               slack username for reboot notfications (default "kured")

Reboot Sentinel File & Period

By default kured checks for the existence of /var/run/reboot-required every sixty minutes; you can override these values with --reboot-sentinel and --period. Each replica of the daemon uses a random offset derived from the period on startup so that nodes don't all contend for the lock simultaneously.

Blocking Reboots via Alerts

You may find it desirable to block automatic node reboots when there are active alerts - you can do so by providing the URL of your Prometheus server:

--prometheus-url=http://prometheus.monitoring.svc.cluster.local

By default the presence of any active (pending or firing) alerts will block reboots, however you can ignore specific alerts:

--alert-filter-regexp=^(RebootRequired|AnotherBenignAlert|...$

See the section on Prometheus metrics for an important application of this filter.

Blocking Reboots via Pods

You can also block reboots of an individual node when specific pods are scheduled on it:

--blocking-pod-selector=runtime=long,cost=expensive

Since label selector strings use commas to express logical 'and', you can specify this parameter multiple times for 'or':

--blocking-pod-selector=runtime=long,cost=expensive
--blocking-pod-selector=name=temperamental

In this case, the presence of either an (appropriately labelled) expensive long running job or a known temperamental pod on a node will stop it rebooting.

Try not to abuse this mechanism - it's better to strive for restartability where possible. If you do use it, make sure you set up a RebootRequired alert as described in the next section so that you can intervene manually if reboots are blocked for too long.

Prometheus Metrics

Each kured pod exposes a single gauge metric (:8080/metrics) that indicates the presence of the sentinel file:

# HELP kured_reboot_required OS requires reboot due to software updates.
# TYPE kured_reboot_required gauge
kured_reboot_required{node="ip-xxx-xxx-xxx-xxx.ec2.internal"} 0

The purpose of this metric is to power an alert which will summon an operator if the cluster cannot reboot itself automatically for a prolonged period:

# Alert if a reboot is required for any machines. Acts as a failsafe for the
# reboot daemon, which will not reboot nodes if there are pending alerts save
# this one.
ALERT RebootRequired
  IF          max(kured_reboot_required) != 0
  FOR         24h
  LABELS      { severity="warning" }
  ANNOTATIONS {
    summary = "Machine(s) require being rebooted, and the reboot daemon has failed to do so for 24 hours",
    impact = "Cluster nodes more vulnerable to security exploits. Eventually, no disk space left.",
    description = "Machine(s) require being rebooted, probably due to kernel update.",
  }

If you choose to employ such an alert and have configured kured to probe for active alerts before rebooting, be sure to specify --alert-filter-regexp=^RebootRequired$ to avoid deadlock!

Slack Notifications

If you specify a Slack hook via --slack-hook-url, kured will notify you immediately prior to rebooting a node:

We recommend setting --slack-username to be the name of the environment, e.g. dev or prod.

Overriding Lock Configuration

The --ds-name and --ds-namespace arguments should match the name and namespace of the daemonset used to deploy the reboot daemon - the locking is implemented by means of an annotation on this resource. The defaults match the daemonset YAML provided in the repository.

Similarly --lock-annotation can be used to change the name of the annotation kured will use to store the lock, but the default is almost certainly safe.

Operation

The example commands in this section assume that you have not overriden the default lock annotation, daemonset name or namespace; if you have, you will have to adjust the commands accordingly.

Testing

You can test your configuration by provoking a reboot on a node:

sudo touch /var/run/reboot-required

Disabling Reboots

If you need to temporarily stop kured from rebooting any nodes, you can take the lock manually:

kubectl -n kube-system annotate ds kured weave.works/kured-node-lock='{"nodeID":"manual"}'

Don't forget to release it afterwards!

Manual Unlock

In exceptional circumstances, such as a node experiencing a permanent failure whilst rebooting, manual intervention may be required to remove the cluster lock:

kubectl -n kube-system annotate ds kured weave.works/kured-node-lock-

NB the - at the end of the command is important - it instructs kubectl to remove that annotation entirely.

Building

See the CircleCI config for the preferred version of Golang. Kured now uses Go Modules, so build instructions vary depending on where you have checked out the repository:

Building outside $GOPATH:

make

Building inside $GOPATH:

GO111MODULE=on make

Frequently Asked/Anticipated Questions

Why is there no latest tag on Docker Hub?

Use of latest for production deployments is bad practice - see here for details. The manifest on master refers to latest for local development testing with minikube only; for production use choose a versioned manifest from the release page.

Getting Help

If you have any questions about, feedback for or problems with kured:

• Invite yourself to the Weave Users Slack.
• Ask a question on the #general slack channel.
• File an issue.

Your feedback is always welcome!