Introduction

DaemonSet is a declarative interface to run selected pods on all ready nodes.

API group Resource Kube Skeleton
extensions/v1beta1 DaemonSet skel
apps/v1beta2 DaemonSet skel

Here's an example Kubernetes DaemonSet:

apiVersion: apps/v1beta2 # for versions before 1.8.0 use extensions/v1beta1
kind: DaemonSet
metadata:
  name: fluentd-elasticsearch
  namespace: kube-system
  labels:
    k8s-app: fluentd-logging
spec:
  selector:
    matchLabels:
      name: fluentd-elasticsearch
  template:
    metadata:
      labels:
        name: fluentd-elasticsearch
    spec:
      tolerations:
      - key: node-role.kubernetes.io/master
        effect: NoSchedule
      containers:
      - name: fluentd-elasticsearch
        image: gcr.io/google-containers/fluentd-elasticsearch:1.20
        resources:
          limits:
            memory: 200Mi
          requests:
            cpu: 100m
            memory: 200Mi
        volumeMounts:
        - name: varlog
          mountPath: /var/log
        - name: varlibdockercontainers
          mountPath: /var/lib/docker/containers
          readOnly: true
      terminationGracePeriodSeconds: 30
      volumes:
      - name: varlog
        hostPath:
          path: /var/log
      - name: varlibdockercontainers
        hostPath:
          path: /var/lib/docker/containers

The following sections contain detailed information about each field in Short syntax, including how the field translates to and from Kubernetes syntax.

API Overview

Field Type K8s counterpart(s) Description
version string apiVersion The version of the resource object
cluster string metadata.clusterName The name of the cluster on which this DaemonSet is running
name string metadata.name The name of the DaemonSet
namespace string metadata.namespace The K8s namespace this DaemonSet will be a member of
labels string metadata.labels Metadata about the DaemonSet, including identifying information
annotations string metadata.annotations Non-identifying information about the DaemonSet
on_delete bool strategy The strategy for performing upgrades. If on_delete is set to true, then the strategy is OnDelete. If not, by default the strategy is RollingUpdate
max_unavailable int or string strategy.rollingUpdate Maximum number of pods that can be unavailable during update. More information below
min_ready int32 minReadySeconds Minimum number of seconds that your pod should be ready before it is considered available
max_revs int32 revisionHistoryLimit Number of old replica sets to retain to allow rollback
selector map[string]string or string selector An expression (string) or a set of key, value pairs (map) that is used to select a set of pods to manage using the daemon-set controller. See Selector Overview
pod_meta TemplateMetadata template Metadata of the Pod that is selected by this DaemonSet. See Template Metadata
volumes Volume spec.volumes Denotes the volumes that are a part of the Pod. See Volume Overview
affinity []Affinity spec.affinity and spec.NodeSelector The Pod's scheduling rules, expressed as (anti-)affinities for nodes or other Pods. See Affinity Overview
node string spec.nodeName Request that the Pod be scheduled on a specific node.
containers Container spec.containers and status Containers that run as a part of the Pod. See Container Overview
init_containers Container spec.initContainers and status Containers that run as a part of the initialization process of the Pod. See Container Overview
dns_policy DNSPolicy spec.dnsPolicy The DNS Policy of the Pod. See DNS Policy Overview
host_aliases []string spec.aliases Set of additional records to be placed in /etc/hosts file inside the Pod. See Host Aliases Overview
host_mode []string spec.hostPID, spec.hostNetwork and spec.hostIPC The Pod's access to host resources. See Host Mode Conversion
hostname string spec.hostname and spec.subDomain The fully qualified domain name of the pod
registry_secrets []string spec.ImagePullSecrets A list of k8s secret resource names that contain credentials to required to access private registries.
restart_policy RestartPolicy spec.restartPolicy Behavior of a Pod when it dies. Can be "always", "on-failure" or "never"
scheduler_name string spec.schedulerName The value from spec.schedulerName is stored here
account string spec.serviceAccountName and automountService AccountToken The Pod's access to the K8s API. See Account Conversion
tolerations []Toleration spec.tolerations Set of host taints this Pod tolerates. See Toleration Conversion
termination_ grace_period int64 spec.termination GracePeriodSeconds Number of seconds to wait before forcefully killing the Pod.
active_deadline int64 spec. activeDeadlineSeconds Number of seconds the Pod is allowed to be active
priority Priority spec.priorityClassName and spec.priority Specifies the Pod's Priority. See Priority
condition []Pod Condition status.conditions The list of current and previous conditions of the Pod. See Pod Condition
node_ip string status.hostIP The IP address of the Pod's host
ip string status.podIP The IP address of the Pod
start_time time status.startTime When the Pod started running
msg string status.message A human readable message explaining Pod's current condition
phase string status.phase The current phase of the Pod
reason string status.reason Reason indicating the cause for the current state of the Pod
qos string status.qosClass The QOS class assigned to the Pod based on resource requirements
fs_gid int64 spec.securityContext. fsGroup Special supplemental group that applies to all the Containers in the Pod
gids []int64 spec.securityContext. supplementalGroups A list of groups applied to the first process in each of the Containers in the Pod

max_unavailable is used to configure RollingUpdate DaemonSet strategy. max_unavailable indicates the maximum number of pods that can be unavailable during update. The value can be number or a percentage value of the total number of replicas. Percentage values are represented using a % symbol at the end of the value.

max_unavailable: 30%

Selector Overview

Selector can be a map value or a string value. If it is a string value, then it can be an expression of type

Valid Operators are

Operator Syntax Description
Eq = Key should be equal to value
Exists N/A Key should exist
NotExists N/A Key should not exist
In = Key should be one of the comma separated values
NotIn != Key should not be one of the comma separated values

Here are valid examples of all the expression operators

selector: key=value # key should be equal to value
selector: key # key should exist
selector: !key # key should not exist
selector: key=value1,value2 # key's value can be any of value1 or value2
selector: key!=value1,value2 # key's value cannot be any of value1 or value2
selector: key&key!=value # composite expression

Note that multiple expressions can be combined using the & symbol

If the selector is a map, then the values in the map are expected to match directly with the labels of a pod.

Template Metadata

Field Type K8s counterpart(s) Description
cluster string metadata.clusterName The name of the cluster on which this Pod is running
name string metadata.name The name of the Pod
namespace string metadata.namespace The K8s namespace this Pod will be a member of
labels string metadata.labels Metadata that could be identifying information about the Pod
annotations string metadata.annotations Non identifying information about the Pod

Examples

daemon_set:
  containers:
  - expose:
    - 80
    image: nginx:1.7.9
    name: nginx
  name: nginx-daemonset
  selector:
    app: nginx
  version: apps/v1beta2
daemon_set:
  containers:
  - expose:
    - 3001
    image: gcr.io/my-project/foo-service:3adfa3e
    name: foo-web
  labels:
    app: foo
  name: foo-web
  replace_on_delete: true   # if this is not set, then default of Rolling Update strategy is used
  selector:
    app: foo
    tier: backend
  version: extensions/v1beta1
daemon_set:
  containers:
  - expose:
    - 80
    image: nginx:1.7.9
    name: nginx
  labels:
    app: nginx
  name: nginx-daemon_set
  selector: app=nginx,haproxy  # string selector (expression)
  version: apps/v1beta2

Skeleton

Short Type Skeleton
DaemonSet skel

Here's a starter skeleton of a Short DaemonSet.

daemon_set:
  containers:
  - cpu:
      min: 100m
    image: gcr.io/google-containers/fluentd-elasticsearch:1.20
    mem:
      max: 200Mi
      min: 200Mi
    name: fluentd-elasticsearch
    volume:
    - mount: /var/log
      store: varlog
    - mount: /var/lib/docker/containers
      store: varlibdockercontainers:ro
  labels:
    k8s-app: fluentd-logging
  name: fluentd-elasticsearch
  namespace: kube-system
  selector:
    name: fluentd-elasticsearch
  termination_grace_period: 30
  version: apps/v1beta2
  volumes:
    varlibdockercontainers: host_path:/var/lib/docker/containers
    varlog: host_path:/var/log