Kubernetes DaemonSet

DaemonSets

• replicates a single pod on all or a subset of nodes in the cluster
• land an agent or daemon on every node for, say, logging, monitoring, etc.

ReplicaSet and DaemonSet both create pods which are expected to be long-running services & try to match the observed and desired state in the cluster.

Use a ReplicaSet when:

• the app is completely decoupled from the node
• multiple copies of the pod can be run on one node
• no scheduling restrictions need to be in place to ensure replicas don't run on the same node

Use a DaemonSet when:

• one pod per node or subset of nodes in the cluster

DaemonSet Scheduler

• By default the a copy is created on every pod
• The nodes can be limited using node selectors which matches to a set of labels
• DaemonSets determine which node a pod will run on when creating it using the nodeName field
• Hence, the k8s scheduler ignores the pods created by DaemonSets

The DaemonSet controller:

• creates a pod on each node that doesn't have one
• if a new node is added to the cluster, the DaemonSet controller adds a pod to it too
• it tries to reconcile the observed state and the desired state

Creating DaemonSets

• the name should be unique across a namespace
• includes a pod spec
• creates the pod on every node if a node selector isn't specified

Limiting DaemonSets to Specific Nodes

1. Add labels to nodes

$ kubectl label nodes k0-default-pool-35609c18-z7tb ssd=true

node "k0-default-pool-35609c18-z7tb" labeled

2. Add the nodeSelector key in the pod spec to limit the number of nodes the pod will run on:

apiVersion: extensions/v1beta1

kind: "DaemonSet"

metadata:

  labels:

    app: nginx

    ssd: "true"

  name: nginx-fast-storage

spec:

  template:

    metadata:

      labels:

        app: nginx

        ssd: "true"

    spec:

      nodeSelector:

        ssd: "true"

      containers:

        - name: nginx

          image: nginx:1.10.0

• If the labels specified as a nodeSelector is added to a new to existing node, the pod will be created on that node. Similarly, if the label is removed from an existing node, the pod will also be removed.

Updating a DaemonSet

• For k8s < 1.6, the DaemonSet was updated and the pods were deleted manually
• For k8s >= 1.6, the DaemonSet has an equivalent to the Deployment object which manages the rollout

Updating a DaemonSet by Deleting Individual Pods

• manually delete the pods associated with a DaemonSet
• delete the entire DaemonSet and create a new one with the updated config. This approach causes downtime as all the pods associated with the DaemonSet are also deleted.

Rolling Update of a DaemonSet

• for backwards compatibility, the default update strategy is the one described above
• to use a rolling update set the following in your yaml: spec.updateStrategy.type: RollingUpdate
• any change to spec.template or sub-field will initiate a rolling update
• it's controlled by the following two params:
• spec.minReadySeconds, which determines how long a Pod must be “ready” before the rolling update proceeds to upgrade subsequent Pods
• spec.updateStrategy.rollingUpdate.maxUnavailable, which indicates how many Pods may be simultaneously updated by the rolling update.

A higher value for spec.updateStrategy.rollingUpdate.maxUnavailable increases the blast radius in case a failure happens but increase the time it takes for the rollout.

Note: In a rolling update, the pods associated with the DaemonSet are upgraded gradually while some pods are still running the old configuration until all pods have the new configuration.

To check the status of the rollout, run: kubectl rollout

Deleting a DaemonSet

kubectl delete -f daemonset.yml

• Deleting a DaemonSet deletes all the pods associated with it
• in order to retain the pods and delete only the DaemonSet, use the flag: --cascade = false