1 - StatefulSet Basics

This tutorial provides an introduction to managing applications with StatefulSets. It demonstrates how to create, delete, scale, and update the Pods of StatefulSets.

Before you begin

Before you begin this tutorial, you should familiarize yourself with the following Kubernetes concepts:

You need to have a Kubernetes cluster, and the kubectl command-line tool must be configured to communicate with your cluster. It is recommended to run this tutorial on a cluster with at least two nodes that are not acting as control plane hosts. If you do not already have a cluster, you can create one by using minikube or you can use one of these Kubernetes playgrounds:

You should configure kubectl to use a context that uses the default namespace. If you are using an existing cluster, make sure that it's OK to use that cluster's default namespace to practice. Ideally, practice in a cluster that doesn't run any real workloads.

It's also useful to read the concept page about StatefulSets.

Objectives

StatefulSets are intended to be used with stateful applications and distributed systems. However, the administration of stateful applications and distributed systems on Kubernetes is a broad, complex topic. In order to demonstrate the basic features of a StatefulSet, and not to conflate the former topic with the latter, you will deploy a simple web application using a StatefulSet.

After this tutorial, you will be familiar with the following.

  • How to create a StatefulSet
  • How a StatefulSet manages its Pods
  • How to delete a StatefulSet
  • How to scale a StatefulSet
  • How to update a StatefulSet's Pods

Creating a StatefulSet

Begin by creating a StatefulSet (and the Service that it relies upon) using the example below. It is similar to the example presented in the StatefulSets concept. It creates a headless Service, nginx, to publish the IP addresses of Pods in the StatefulSet, web.

apiVersion: v1
kind: Service
metadata:
  name: nginx
  labels:
    app: nginx
spec:
  ports:
  - port: 80
    name: web
  clusterIP: None
  selector:
    app: nginx
---
apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: web
spec:
  serviceName: "nginx"
  replicas: 2
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
      - name: nginx
        image: registry.k8s.io/nginx-slim:0.8
        ports:
        - containerPort: 80
          name: web
        volumeMounts:
        - name: www
          mountPath: /usr/share/nginx/html
  volumeClaimTemplates:
  - metadata:
      name: www
    spec:
      accessModes: [ "ReadWriteOnce" ]
      resources:
        requests:
          storage: 1Gi

You will need to use at least two terminal windows. In the first terminal, use kubectl get to watch the creation of the StatefulSet's Pods.

# use this terminal to run commands that specify --watch
# end this watch when you are asked to start a new watch
kubectl get pods --watch -l app=nginx

In the second terminal, use kubectl apply to create the headless Service and StatefulSet:

kubectl apply -f https://k8s.io/examples/application/web/web.yaml
service/nginx created
statefulset.apps/web created

The command above creates two Pods, each running an NGINX webserver. Get the nginx Service...

kubectl get service nginx
NAME      TYPE         CLUSTER-IP   EXTERNAL-IP   PORT(S)   AGE
nginx     ClusterIP    None         <none>        80/TCP    12s

...then get the web StatefulSet, to verify that both were created successfully:

kubectl get statefulset web
NAME      DESIRED   CURRENT   AGE
web       2         1         20s

Ordered Pod creation

A StatefulSet defaults to creating its Pods in a strict order.

For a StatefulSet with n replicas, when Pods are being deployed, they are created sequentially, ordered from {0..n-1}. Examine the output of the kubectl get command in the first terminal. Eventually, the output will look like the example below.

# Do not start a new watch;
# this should already be running
kubectl get pods --watch -l app=nginx
NAME      READY     STATUS    RESTARTS   AGE
web-0     0/1       Pending   0          0s
web-0     0/1       Pending   0         0s
web-0     0/1       ContainerCreating   0         0s
web-0     1/1       Running   0         19s
web-1     0/1       Pending   0         0s
web-1     0/1       Pending   0         0s
web-1     0/1       ContainerCreating   0         0s
web-1     1/1       Running   0         18s

Notice that the web-1 Pod is not launched until the web-0 Pod is Running (see Pod Phase) and Ready (see type in Pod Conditions).

Later in this tutorial you will practice parallel startup.

Pods in a StatefulSet

Pods in a StatefulSet have a unique ordinal index and a stable network identity.

Examining the Pod's ordinal index

Get the StatefulSet's Pods:

kubectl get pods -l app=nginx
NAME      READY     STATUS    RESTARTS   AGE
web-0     1/1       Running   0          1m
web-1     1/1       Running   0          1m

As mentioned in the StatefulSets concept, the Pods in a StatefulSet have a sticky, unique identity. This identity is based on a unique ordinal index that is assigned to each Pod by the StatefulSet controller.
The Pods' names take the form <statefulset name>-<ordinal index>. Since the web StatefulSet has two replicas, it creates two Pods, web-0 and web-1.

Using stable network identities

Each Pod has a stable hostname based on its ordinal index. Use kubectl exec to execute the hostname command in each Pod:

for i in 0 1; do kubectl exec "web-$i" -- sh -c 'hostname'; done
web-0
web-1

Use kubectl run to execute a container that provides the nslookup command from the dnsutils package. Using nslookup on the Pods' hostnames, you can examine their in-cluster DNS addresses:

kubectl run -i --tty --image busybox:1.28 dns-test --restart=Never --rm

which starts a new shell. In that new shell, run:

# Run this in the dns-test container shell
nslookup web-0.nginx

The output is similar to:

Server:    10.0.0.10
Address 1: 10.0.0.10 kube-dns.kube-system.svc.cluster.local

Name:      web-0.nginx
Address 1: 10.244.1.6

nslookup web-1.nginx
Server:    10.0.0.10
Address 1: 10.0.0.10 kube-dns.kube-system.svc.cluster.local

Name:      web-1.nginx
Address 1: 10.244.2.6

(and now exit the container shell: exit)

The CNAME of the headless service points to SRV records (one for each Pod that is Running and Ready). The SRV records point to A record entries that contain the Pods' IP addresses.

In one terminal, watch the StatefulSet's Pods:

# Start a new watch
# End this watch when you've seen that the delete is finished
kubectl get pod --watch -l app=nginx

In a second terminal, use kubectl delete to delete all the Pods in the StatefulSet:

kubectl delete pod -l app=nginx
pod "web-0" deleted
pod "web-1" deleted

Wait for the StatefulSet to restart them, and for both Pods to transition to Running and Ready:

# This should already be running
kubectl get pod --watch -l app=nginx
NAME      READY     STATUS              RESTARTS   AGE
web-0     0/1       ContainerCreating   0          0s
NAME      READY     STATUS    RESTARTS   AGE
web-0     1/1       Running   0          2s
web-1     0/1       Pending   0         0s
web-1     0/1       Pending   0         0s
web-1     0/1       ContainerCreating   0         0s
web-1     1/1       Running   0         34s

Use kubectl exec and kubectl run to view the Pods' hostnames and in-cluster DNS entries. First, view the Pods' hostnames:

for i in 0 1; do kubectl exec web-$i -- sh -c 'hostname'; done
web-0
web-1

then, run:

kubectl run -i --tty --image busybox:1.28 dns-test --restart=Never --rm

which starts a new shell.
In that new shell, run:

# Run this in the dns-test container shell
nslookup web-0.nginx

The output is similar to:

Server:    10.0.0.10
Address 1: 10.0.0.10 kube-dns.kube-system.svc.cluster.local

Name:      web-0.nginx
Address 1: 10.244.1.7

nslookup web-1.nginx
Server:    10.0.0.10
Address 1: 10.0.0.10 kube-dns.kube-system.svc.cluster.local

Name:      web-1.nginx
Address 1: 10.244.2.8

(and now exit the container shell: exit)

The Pods' ordinals, hostnames, SRV records, and A record names have not changed, but the IP addresses associated with the Pods may have changed. In the cluster used for this tutorial, they have. This is why it is important not to configure other applications to connect to Pods in a StatefulSet by the IP address of a particular Pod (it is OK to connect to Pods by resolving their hostname).

Discovery for specific Pods in a StatefulSet

If you need to find and connect to the active members of a StatefulSet, you should query the CNAME of the headless Service (nginx.default.svc.cluster.local). The SRV records associated with the CNAME will contain only the Pods in the StatefulSet that are Running and Ready.

If your application already implements connection logic that tests for liveness and readiness, you can use the SRV records of the Pods ( web-0.nginx.default.svc.cluster.local, web-1.nginx.default.svc.cluster.local), as they are stable, and your application will be able to discover the Pods' addresses when they transition to Running and Ready.

If your application wants to find any healthy Pod in a StatefulSet, and therefore does not need to track each specific Pod, you could also connect to the IP address of a type: ClusterIP Service, backed by the Pods in that StatefulSet. You can use the same Service that tracks the StatefulSet (specified in the serviceName of the StatefulSet) or a separate Service that selects the right set of Pods.

Writing to stable storage

Get the PersistentVolumeClaims for web-0 and web-1:

kubectl get pvc -l app=nginx

The output is similar to:

NAME        STATUS    VOLUME                                     CAPACITY   ACCESSMODES   AGE
www-web-0   Bound     pvc-15c268c7-b507-11e6-932f-42010a800002   1Gi        RWO           48s
www-web-1   Bound     pvc-15c79307-b507-11e6-932f-42010a800002   1Gi        RWO           48s

The StatefulSet controller created two PersistentVolumeClaims that are bound to two PersistentVolumes.

As the cluster used in this tutorial is configured to dynamically provision PersistentVolumes, the PersistentVolumes were created and bound automatically.

The NGINX webserver, by default, serves an index file from /usr/share/nginx/html/index.html. The volumeMounts field in the StatefulSet's spec ensures that the /usr/share/nginx/html directory is backed by a PersistentVolume.

Write the Pods' hostnames to their index.html files and verify that the NGINX webservers serve the hostnames:

for i in 0 1; do kubectl exec "web-$i" -- sh -c 'echo "$(hostname)" > /usr/share/nginx/html/index.html'; done

for i in 0 1; do kubectl exec -i -t "web-$i" -- curl http://localhost/; done
web-0
web-1

In one terminal, watch the StatefulSet's Pods:

# End this watch when you've reached the end of the section.
# At the start of "Scaling a StatefulSet" you'll start a new watch.
kubectl get pod --watch -l app=nginx

In a second terminal, delete all of the StatefulSet's Pods:

kubectl delete pod -l app=nginx
pod "web-0" deleted
pod "web-1" deleted

Examine the output of the kubectl get command in the first terminal, and wait for all of the Pods to transition to Running and Ready.

# This should already be running
kubectl get pod --watch -l app=nginx
NAME      READY     STATUS              RESTARTS   AGE
web-0     0/1       ContainerCreating   0          0s
NAME      READY     STATUS    RESTARTS   AGE
web-0     1/1       Running   0          2s
web-1     0/1       Pending   0         0s
web-1     0/1       Pending   0         0s
web-1     0/1       ContainerCreating   0         0s
web-1     1/1       Running   0         34s

Verify the web servers continue to serve their hostnames:

for i in 0 1; do kubectl exec -i -t "web-$i" -- curl http://localhost/; done
web-0
web-1

Even though web-0 and web-1 were rescheduled, they continue to serve their hostnames because the PersistentVolumes associated with their PersistentVolumeClaims are remounted to their volumeMounts. No matter what node web-0and web-1 are scheduled on, their PersistentVolumes will be mounted to the appropriate mount points.

Scaling a StatefulSet

Scaling a StatefulSet refers to increasing or decreasing the number of replicas (horizontal scaling). This is accomplished by updating the replicas field. You can use either kubectl scale or kubectl patch to scale a StatefulSet.

Scaling up

Scaling up means adding more replicas. Provided that your app is able to distribute work across the StatefulSet, the new larger set of Pods can perform more of that work.

In one terminal window, watch the Pods in the StatefulSet:

# If you already have a watch running, you can continue using that.
# Otherwise, start one.
# End this watch when there are 5 healthy Pods for the StatefulSet
kubectl get pods --watch -l app=nginx

In another terminal window, use kubectl scale to scale the number of replicas to 5:

kubectl scale sts web --replicas=5
statefulset.apps/web scaled

Examine the output of the kubectl get command in the first terminal, and wait for the three additional Pods to transition to Running and Ready.

# This should already be running
kubectl get pod --watch -l app=nginx
NAME      READY     STATUS    RESTARTS   AGE
web-0     1/1       Running   0          2h
web-1     1/1       Running   0          2h
NAME      READY     STATUS    RESTARTS   AGE
web-2     0/1       Pending   0          0s
web-2     0/1       Pending   0         0s
web-2     0/1       ContainerCreating   0         0s
web-2     1/1       Running   0         19s
web-3     0/1       Pending   0         0s
web-3     0/1       Pending   0         0s
web-3     0/1       ContainerCreating   0         0s
web-3     1/1       Running   0         18s
web-4     0/1       Pending   0         0s
web-4     0/1       Pending   0         0s
web-4     0/1       ContainerCreating   0         0s
web-4     1/1       Running   0         19s

The StatefulSet controller scaled the number of replicas. As with StatefulSet creation, the StatefulSet controller created each Pod sequentially with respect to its ordinal index, and it waited for each Pod's predecessor to be Running and Ready before launching the subsequent Pod.

Scaling down

Scaling down means reducing the number of replicas. For example, you might do this because the level of traffic to a service has decreased, and at the current scale there are idle resources.

In one terminal, watch the StatefulSet's Pods:

# End this watch when there are only 3 Pods for the StatefulSet
kubectl get pod --watch -l app=nginx

In another terminal, use kubectl patch to scale the StatefulSet back down to three replicas:

kubectl patch sts web -p '{"spec":{"replicas":3}}'
statefulset.apps/web patched

Wait for web-4 and web-3 to transition to Terminating.

# This should already be running
kubectl get pods --watch -l app=nginx
NAME      READY     STATUS              RESTARTS   AGE
web-0     1/1       Running             0          3h
web-1     1/1       Running             0          3h
web-2     1/1       Running             0          55s
web-3     1/1       Running             0          36s
web-4     0/1       ContainerCreating   0          18s
NAME      READY     STATUS    RESTARTS   AGE
web-4     1/1       Running   0          19s
web-4     1/1       Terminating   0         24s
web-4     1/1       Terminating   0         24s
web-3     1/1       Terminating   0         42s
web-3     1/1       Terminating   0         42s

Ordered Pod termination

The control plane deleted one Pod at a time, in reverse order with respect to its ordinal index, and it waited for each Pod to be completely shut down before deleting the next one.

Get the StatefulSet's PersistentVolumeClaims:

kubectl get pvc -l app=nginx
NAME        STATUS    VOLUME                                     CAPACITY   ACCESSMODES   AGE
www-web-0   Bound     pvc-15c268c7-b507-11e6-932f-42010a800002   1Gi        RWO           13h
www-web-1   Bound     pvc-15c79307-b507-11e6-932f-42010a800002   1Gi        RWO           13h
www-web-2   Bound     pvc-e1125b27-b508-11e6-932f-42010a800002   1Gi        RWO           13h
www-web-3   Bound     pvc-e1176df6-b508-11e6-932f-42010a800002   1Gi        RWO           13h
www-web-4   Bound     pvc-e11bb5f8-b508-11e6-932f-42010a800002   1Gi        RWO           13h

There are still five PersistentVolumeClaims and five PersistentVolumes. When exploring a Pod's stable storage, you saw that the PersistentVolumes mounted to the Pods of a StatefulSet are not deleted when the StatefulSet's Pods are deleted. This is still true when Pod deletion is caused by scaling the StatefulSet down.

Updating StatefulSets

The StatefulSet controller supports automated updates. The strategy used is determined by the spec.updateStrategy field of the StatefulSet API object. This feature can be used to upgrade the container images, resource requests and/or limits, labels, and annotations of the Pods in a StatefulSet.

There are two valid update strategies, RollingUpdate (the default) and OnDelete.

RollingUpdate

The RollingUpdate update strategy will update all Pods in a StatefulSet, in reverse ordinal order, while respecting the StatefulSet guarantees.

You can split updates to a StatefulSet that uses the RollingUpdate strategy into partitions, by specifying .spec.updateStrategy.rollingUpdate.partition. You'll practice that later in this tutorial.

First, try a simple rolling update.

In one terminal window, patch the web StatefulSet to change the container image again:

kubectl patch statefulset web --type='json' -p='[{"op": "replace", "path": "/spec/template/spec/containers/0/image", "value":"gcr.io/google_containers/nginx-slim:0.8"}]'
statefulset.apps/web patched

In another terminal, watch the Pods in the StatefulSet:

# End this watch when the rollout is complete
#
# If you're not sure, leave it running one more minute
kubectl get pod -l app=nginx --watch

The output is similar to:

NAME      READY     STATUS    RESTARTS   AGE
web-0     1/1       Running   0          7m
web-1     1/1       Running   0          7m
web-2     1/1       Running   0          8m
web-2     1/1       Terminating   0         8m
web-2     1/1       Terminating   0         8m
web-2     0/1       Terminating   0         8m
web-2     0/1       Terminating   0         8m
web-2     0/1       Terminating   0         8m
web-2     0/1       Terminating   0         8m
web-2     0/1       Pending   0         0s
web-2     0/1       Pending   0         0s
web-2     0/1       ContainerCreating   0         0s
web-2     1/1       Running   0         19s
web-1     1/1       Terminating   0         8m
web-1     0/1       Terminating   0         8m
web-1     0/1       Terminating   0         8m
web-1     0/1       Terminating   0         8m
web-1     0/1       Pending   0         0s
web-1     0/1       Pending   0         0s
web-1     0/1       ContainerCreating   0         0s
web-1     1/1       Running   0         6s
web-0     1/1       Terminating   0         7m
web-0     1/1       Terminating   0         7m
web-0     0/1       Terminating   0         7m
web-0     0/1       Terminating   0         7m
web-0     0/1       Terminating   0         7m
web-0     0/1       Terminating   0         7m
web-0     0/1       Pending   0         0s
web-0     0/1       Pending   0         0s
web-0     0/1       ContainerCreating   0         0s
web-0     1/1       Running   0         10s

The Pods in the StatefulSet are updated in reverse ordinal order. The StatefulSet controller terminates each Pod, and waits for it to transition to Running and Ready prior to updating the next Pod. Note that, even though the StatefulSet controller will not proceed to update the next Pod until its ordinal successor is Running and Ready, it will restore any Pod that fails during the update to that Pod's existing version.

Pods that have already received the update will be restored to the updated version, and Pods that have not yet received the update will be restored to the previous version. In this way, the controller attempts to continue to keep the application healthy and the update consistent in the presence of intermittent failures.

Get the Pods to view their container images:

for p in 0 1 2; do kubectl get pod "web-$p" --template '{{range $i, $c := .spec.containers}}{{$c.image}}{{end}}'; echo; done
registry.k8s.io/nginx-slim:0.8
registry.k8s.io/nginx-slim:0.8
registry.k8s.io/nginx-slim:0.8

All the Pods in the StatefulSet are now running the previous container image.

Staging an update

You can split updates to a StatefulSet that uses the RollingUpdate strategy into partitions, by specifying .spec.updateStrategy.rollingUpdate.partition.

For more context, you can read Partitioned rolling updates in the StatefulSet concept page.

You can stage an update to a StatefulSet by using the partition field within .spec.updateStrategy.rollingUpdate. For this update, you will keep the existing Pods in the StatefulSet unchanged whilst you change the pod template for the StatefulSet. Then you - or, outside of a tutorial, some external automation - can trigger that prepared update.

First, patch the web StatefulSet to add a partition to the updateStrategy field:

# The value of "partition" determines which ordinals a change applies to
# Make sure to use a number bigger than the last ordinal for the
# StatefulSet
kubectl patch statefulset web -p '{"spec":{"updateStrategy":{"type":"RollingUpdate","rollingUpdate":{"partition":3}}}}'
statefulset.apps/web patched

Patch the StatefulSet again to change the container image that this StatefulSet uses:

kubectl patch statefulset web --type='json' -p='[{"op": "replace", "path": "/spec/template/spec/containers/0/image", "value":"registry.k8s.io/nginx-slim:0.7"}]'
statefulset.apps/web patched

Delete a Pod in the StatefulSet:

kubectl delete pod web-2
pod "web-2" deleted

Wait for the replacement web-2 Pod to be Running and Ready:

# End the watch when you see that web-2 is healthy
kubectl get pod -l app=nginx --watch
NAME      READY     STATUS              RESTARTS   AGE
web-0     1/1       Running             0          4m
web-1     1/1       Running             0          4m
web-2     0/1       ContainerCreating   0          11s
web-2     1/1       Running   0         18s

Get the Pod's container image:

kubectl get pod web-2 --template '{{range $i, $c := .spec.containers}}{{$c.image}}{{end}}'
registry.k8s.io/nginx-slim:0.8

Notice that, even though the update strategy is RollingUpdate the StatefulSet restored the Pod with the original container image. This is because the ordinal of the Pod is less than the partition specified by the updateStrategy.

Rolling out a canary

You're now going to try a canary rollout of that staged change.

You can roll out a canary (to test the modified template) by decrementing the partition you specified above.

Patch the StatefulSet to decrement the partition:

# The value of "partition" should match the highest existing ordinal for
# the StatefulSet
kubectl patch statefulset web -p '{"spec":{"updateStrategy":{"type":"RollingUpdate","rollingUpdate":{"partition":2}}}}'
statefulset.apps/web patched

The control plane triggers replacement for web-2 (implemented by a graceful delete followed by creating a new Pod once the deletion is complete). Wait for the new web-2 Pod to be Running and Ready.

# This should already be running
kubectl get pod -l app=nginx --watch
NAME      READY     STATUS              RESTARTS   AGE
web-0     1/1       Running             0          4m
web-1     1/1       Running             0          4m
web-2     0/1       ContainerCreating   0          11s
web-2     1/1       Running   0         18s

Get the Pod's container:

kubectl get pod web-2 --template '{{range $i, $c := .spec.containers}}{{$c.image}}{{end}}'
registry.k8s.io/nginx-slim:0.7

When you changed the partition, the StatefulSet controller automatically updated the web-2 Pod because the Pod's ordinal was greater than or equal to the partition.

Delete the web-1 Pod:

kubectl delete pod web-1
pod "web-1" deleted

Wait for the web-1 Pod to be Running and Ready.

# This should already be running
kubectl get pod -l app=nginx --watch

The output is similar to:

NAME      READY     STATUS        RESTARTS   AGE
web-0     1/1       Running       0          6m
web-1     0/1       Terminating   0          6m
web-2     1/1       Running       0          2m
web-1     0/1       Terminating   0         6m
web-1     0/1       Terminating   0         6m
web-1     0/1       Terminating   0         6m
web-1     0/1       Pending   0         0s
web-1     0/1       Pending   0         0s
web-1     0/1       ContainerCreating   0         0s
web-1     1/1       Running   0         18s

Get the web-1 Pod's container image:

kubectl get pod web-1 --template '{{range $i, $c := .spec.containers}}{{$c.image}}{{end}}'
registry.k8s.io/nginx-slim:0.8

web-1 was restored to its original configuration because the Pod's ordinal was less than the partition. When a partition is specified, all Pods with an ordinal that is greater than or equal to the partition will be updated when the StatefulSet's .spec.template is updated. If a Pod that has an ordinal less than the partition is deleted or otherwise terminated, it will be restored to its original configuration.

Phased roll outs

You can perform a phased roll out (e.g. a linear, geometric, or exponential roll out) using a partitioned rolling update in a similar manner to how you rolled out a canary. To perform a phased roll out, set the partition to the ordinal at which you want the controller to pause the update.

The partition is currently set to 2. Set the partition to 0:

kubectl patch statefulset web -p '{"spec":{"updateStrategy":{"type":"RollingUpdate","rollingUpdate":{"partition":0}}}}'
statefulset.apps/web patched

Wait for all of the Pods in the StatefulSet to become Running and Ready.

# This should already be running
kubectl get pod -l app=nginx --watch

The output is similar to:

NAME      READY     STATUS              RESTARTS   AGE
web-0     1/1       Running             0          3m
web-1     0/1       ContainerCreating   0          11s
web-2     1/1       Running             0          2m
web-1     1/1       Running   0         18s
web-0     1/1       Terminating   0         3m
web-0     1/1       Terminating   0         3m
web-0     0/1       Terminating   0         3m
web-0     0/1       Terminating   0         3m
web-0     0/1       Terminating   0         3m
web-0     0/1       Terminating   0         3m
web-0     0/1       Pending   0         0s
web-0     0/1       Pending   0         0s
web-0     0/1       ContainerCreating   0         0s
web-0     1/1       Running   0         3s

Get the container image details for the Pods in the StatefulSet:

for p in 0 1 2; do kubectl get pod "web-$p" --template '{{range $i, $c := .spec.containers}}{{$c.image}}{{end}}'; echo; done
registry.k8s.io/nginx-slim:0.7
registry.k8s.io/nginx-slim:0.7
registry.k8s.io/nginx-slim:0.7

By moving the partition to 0, you allowed the StatefulSet to continue the update process.

OnDelete

You select this update strategy for a StatefulSet by setting the .spec.template.updateStrategy.type to OnDelete.

Patch the web StatefulSet to use the OnDelete update strategy:

kubectl patch statefulset web -p '{"spec":{"updateStrategy":{"type":"OnDelete"}}}'
statefulset.apps/web patched

When you select this update strategy, the StatefulSet controller does not automatically update Pods when a modification is made to the StatefulSet's .spec.template field. You need to manage the rollout yourself - either manually, or using separate automation.

Deleting StatefulSets

StatefulSet supports both non-cascading and cascading deletion. In a non-cascading delete, the StatefulSet's Pods are not deleted when the StatefulSet is deleted. In a cascading delete, both the StatefulSet and its Pods are deleted.

Read Use Cascading Deletion in a Cluster to learn about cascading deletion generally.

Non-cascading delete

In one terminal window, watch the Pods in the StatefulSet.

# End this watch when there are no Pods for the StatefulSet
kubectl get pods --watch -l app=nginx

Use kubectl delete to delete the StatefulSet. Make sure to supply the --cascade=orphan parameter to the command. This parameter tells Kubernetes to only delete the StatefulSet, and to not delete any of its Pods.

kubectl delete statefulset web --cascade=orphan
statefulset.apps "web" deleted

Get the Pods, to examine their status:

kubectl get pods -l app=nginx
NAME      READY     STATUS    RESTARTS   AGE
web-0     1/1       Running   0          6m
web-1     1/1       Running   0          7m
web-2     1/1       Running   0          5m

Even though web has been deleted, all of the Pods are still Running and Ready. Delete web-0:

kubectl delete pod web-0
pod "web-0" deleted

Get the StatefulSet's Pods:

kubectl get pods -l app=nginx
NAME      READY     STATUS    RESTARTS   AGE
web-1     1/1       Running   0          10m
web-2     1/1       Running   0          7m

As the web StatefulSet has been deleted, web-0 has not been relaunched.

In one terminal, watch the StatefulSet's Pods.

# Leave this watch running until the next time you start a watch
kubectl get pods --watch -l app=nginx

In a second terminal, recreate the StatefulSet. Note that, unless you deleted the nginx Service (which you should not have), you will see an error indicating that the Service already exists.

kubectl apply -f https://k8s.io/examples/application/web/web.yaml
statefulset.apps/web created
service/nginx unchanged

Ignore the error. It only indicates that an attempt was made to create the nginx headless Service even though that Service already exists.

Examine the output of the kubectl get command running in the first terminal.

# This should already be running
kubectl get pods --watch -l app=nginx
NAME      READY     STATUS    RESTARTS   AGE
web-1     1/1       Running   0          16m
web-2     1/1       Running   0          2m
NAME      READY     STATUS    RESTARTS   AGE
web-0     0/1       Pending   0          0s
web-0     0/1       Pending   0         0s
web-0     0/1       ContainerCreating   0         0s
web-0     1/1       Running   0         18s
web-2     1/1       Terminating   0         3m
web-2     0/1       Terminating   0         3m
web-2     0/1       Terminating   0         3m
web-2     0/1       Terminating   0         3m

When the web StatefulSet was recreated, it first relaunched web-0. Since web-1 was already Running and Ready, when web-0 transitioned to Running and Ready, it adopted this Pod. Since you recreated the StatefulSet with replicas equal to 2, once web-0 had been recreated, and once web-1 had been determined to already be Running and Ready, web-2 was terminated.

Now take another look at the contents of the index.html file served by the Pods' webservers:

for i in 0 1; do kubectl exec -i -t "web-$i" -- curl http://localhost/; done
web-0
web-1

Even though you deleted both the StatefulSet and the web-0 Pod, it still serves the hostname originally entered into its index.html file. This is because the StatefulSet never deletes the PersistentVolumes associated with a Pod. When you recreated the StatefulSet and it relaunched web-0, its original PersistentVolume was remounted.

Cascading delete

In one terminal window, watch the Pods in the StatefulSet.

# Leave this running until the next page section
kubectl get pods --watch -l app=nginx

In another terminal, delete the StatefulSet again. This time, omit the --cascade=orphan parameter.

kubectl delete statefulset web
statefulset.apps "web" deleted

Examine the output of the kubectl get command running in the first terminal, and wait for all of the Pods to transition to Terminating.

# This should already be running
kubectl get pods --watch -l app=nginx
NAME      READY     STATUS    RESTARTS   AGE
web-0     1/1       Running   0          11m
web-1     1/1       Running   0          27m
NAME      READY     STATUS        RESTARTS   AGE
web-0     1/1       Terminating   0          12m
web-1     1/1       Terminating   0         29m
web-0     0/1       Terminating   0         12m
web-0     0/1       Terminating   0         12m
web-0     0/1       Terminating   0         12m
web-1     0/1       Terminating   0         29m
web-1     0/1       Terminating   0         29m
web-1     0/1       Terminating   0         29m

As you saw in the Scaling Down section, the Pods are terminated one at a time, with respect to the reverse order of their ordinal indices. Before terminating a Pod, the StatefulSet controller waits for the Pod's successor to be completely terminated.

kubectl delete service nginx
service "nginx" deleted

Recreate the StatefulSet and headless Service one more time:

kubectl apply -f https://k8s.io/examples/application/web/web.yaml
service/nginx created
statefulset.apps/web created

When all of the StatefulSet's Pods transition to Running and Ready, retrieve the contents of their index.html files:

for i in 0 1; do kubectl exec -i -t "web-$i" -- curl http://localhost/; done
web-0
web-1

Even though you completely deleted the StatefulSet, and all of its Pods, the Pods are recreated with their PersistentVolumes mounted, and web-0 and web-1 continue to serve their hostnames.

Finally, delete the nginx Service...

kubectl delete service nginx
service "nginx" deleted

...and the web StatefulSet:

kubectl delete statefulset web
statefulset "web" deleted

Pod management policy

For some distributed systems, the StatefulSet ordering guarantees are unnecessary and/or undesirable. These systems require only uniqueness and identity.

You can specify a Pod management policy to avoid this strict ordering; either OrderedReady (the default), or Parallel.

OrderedReady Pod management

OrderedReady pod management is the default for StatefulSets. It tells the StatefulSet controller to respect the ordering guarantees demonstrated above.

Use this when your application requires or expects that changes, such as rolling out a new version of your application, happen in the strict order of the ordinal (pod number) that the StatefulSet provides. In other words, if you have Pods app-0, app-1 and app-2, Kubernetes will update app-0 first and check it. Once the checks are good, Kubernetes updates app-1 and finally app-2.

If you added two more Pods, Kubernetes would set up app-3 and wait for that to become healthy before deploying app-4.

Because this is the default setting, you've already practised using it.

Parallel Pod management

The alternative, Parallel pod management, tells the StatefulSet controller to launch or terminate all Pods in parallel, and not to wait for Pods to become Running and Ready or completely terminated prior to launching or terminating another Pod.

The Parallel pod management option only affects the behavior for scaling operations. Updates are not affected; Kubernetes still rolls out changes in order. For this tutorial, the application is very simple: a webserver that tells you its hostname (because this is a StatefulSet, the hostname for each Pod is different and predictable).

apiVersion: v1
kind: Service
metadata:
  name: nginx
  labels:
    app: nginx
spec:
  ports:
  - port: 80
    name: web
  clusterIP: None
  selector:
    app: nginx
---
apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: web
spec:
  serviceName: "nginx"
  podManagementPolicy: "Parallel"
  replicas: 2
  selector:
    matchLabels:
      app: nginx
  template:
    metadata:
      labels:
        app: nginx
    spec:
      containers:
      - name: nginx
        image: registry.k8s.io/nginx-slim:0.8
        ports:
        - containerPort: 80
          name: web
        volumeMounts:
        - name: www
          mountPath: /usr/share/nginx/html
  volumeClaimTemplates:
  - metadata:
      name: www
    spec:
      accessModes: [ "ReadWriteOnce" ]
      resources:
        requests:
          storage: 1Gi

This manifest is identical to the one you downloaded above except that the .spec.podManagementPolicy of the web StatefulSet is set to Parallel.

In one terminal, watch the Pods in the StatefulSet.

# Leave this watch running until the end of the section
kubectl get pod -l app=nginx --watch

In another terminal, reconfigure the StatefulSet for Parallel Pod management:

kubectl apply -f https://k8s.io/examples/application/web/web-parallel.yaml
service/nginx updated
statefulset.apps/web updated

Keep the terminal open where you're running the watch. In another terminal window, scale the StatefulSet:

kubectl scale statefulset/web --replicas=5
statefulset.apps/web scaled

Examine the output of the terminal where the kubectl get command is running. It may look something like

web-3     0/1       Pending   0         0s
web-3     0/1       Pending   0         0s
web-3     0/1       Pending   0         7s
web-3     0/1       ContainerCreating   0         7s
web-2     0/1       Pending   0         0s
web-4     0/1       Pending   0         0s
web-2     1/1       Running   0         8s
web-4     0/1       ContainerCreating   0         4s
web-3     1/1       Running   0         26s
web-4     1/1       Running   0         2s

The StatefulSet launched three new Pods, and it did not wait for the first to become Running and Ready prior to launching the second and third Pods.

This approach is useful if your workload has a stateful element, or needs Pods to be able to identify each other with predictable naming, and especially if you sometimes need to provide a lot more capacity quickly. If this simple web service for the tutorial suddenly got an extra 1,000,000 requests per minute then you would want to run some more Pods - but you also would not want to wait for each new Pod to launch. Starting the extra Pods in parallel cuts the time between requesting the extra capacity and having it available for use.

Cleaning up

You should have two terminals open, ready for you to run kubectl commands as part of cleanup.

kubectl delete sts web
# sts is an abbreviation for statefulset

You can watch kubectl get to see those Pods being deleted.

# end the watch when you've seen what you need to
kubectl get pod -l app=nginx --watch
web-3     1/1       Terminating   0         9m
web-2     1/1       Terminating   0         9m
web-3     1/1       Terminating   0         9m
web-2     1/1       Terminating   0         9m
web-1     1/1       Terminating   0         44m
web-0     1/1       Terminating   0         44m
web-0     0/1       Terminating   0         44m
web-3     0/1       Terminating   0         9m
web-2     0/1       Terminating   0         9m
web-1     0/1       Terminating   0         44m
web-0     0/1       Terminating   0         44m
web-2     0/1       Terminating   0         9m
web-2     0/1       Terminating   0         9m
web-2     0/1       Terminating   0         9m
web-1     0/1       Terminating   0         44m
web-1     0/1       Terminating   0         44m
web-1     0/1       Terminating   0         44m
web-0     0/1       Terminating   0         44m
web-0     0/1       Terminating   0         44m
web-0     0/1       Terminating   0         44m
web-3     0/1       Terminating   0         9m
web-3     0/1       Terminating   0         9m
web-3     0/1       Terminating   0         9m

During deletion, a StatefulSet removes all Pods concurrently; it does not wait for a Pod's ordinal successor to terminate prior to deleting that Pod.

Close the terminal where the kubectl get command is running and delete the nginx Service:

kubectl delete svc nginx

Delete the persistent storage media for the PersistentVolumes used in this tutorial.

kubectl get pvc
NAME        STATUS   VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS   AGE
www-web-0   Bound    pvc-2bf00408-d366-4a12-bad0-1869c65d0bee   1Gi        RWO            standard       25m
www-web-1   Bound    pvc-ba3bfe9c-413e-4b95-a2c0-3ea8a54dbab4   1Gi        RWO            standard       24m
www-web-2   Bound    pvc-cba6cfa6-3a47-486b-a138-db5930207eaf   1Gi        RWO            standard       15m
www-web-3   Bound    pvc-0c04d7f0-787a-4977-8da3-d9d3a6d8d752   1Gi        RWO            standard       15m
www-web-4   Bound    pvc-b2c73489-e70b-4a4e-9ec1-9eab439aa43e   1Gi        RWO            standard       14m
kubectl get pv
NAME                                       CAPACITY   ACCESS MODES   RECLAIM POLICY   STATUS   CLAIM               STORAGECLASS   REASON   AGE
pvc-0c04d7f0-787a-4977-8da3-d9d3a6d8d752   1Gi        RWO            Delete           Bound    default/www-web-3   standard                15m
pvc-2bf00408-d366-4a12-bad0-1869c65d0bee   1Gi        RWO            Delete           Bound    default/www-web-0   standard                25m
pvc-b2c73489-e70b-4a4e-9ec1-9eab439aa43e   1Gi        RWO            Delete           Bound    default/www-web-4   standard                14m
pvc-ba3bfe9c-413e-4b95-a2c0-3ea8a54dbab4   1Gi        RWO            Delete           Bound    default/www-web-1   standard                24m
pvc-cba6cfa6-3a47-486b-a138-db5930207eaf   1Gi        RWO            Delete           Bound    default/www-web-2   standard                15m
kubectl delete pvc www-web-0 www-web-1 www-web-2 www-web-3 www-web-4
persistentvolumeclaim "www-web-0" deleted
persistentvolumeclaim "www-web-1" deleted
persistentvolumeclaim "www-web-2" deleted
persistentvolumeclaim "www-web-3" deleted
persistentvolumeclaim "www-web-4" deleted
kubectl get pvc
No resources found in default namespace.

2 - Example: Deploying WordPress and MySQL with Persistent Volumes

This tutorial shows you how to deploy a WordPress site and a MySQL database using Minikube. Both applications use PersistentVolumes and PersistentVolumeClaims to store data.

A PersistentVolume (PV) is a piece of storage in the cluster that has been manually provisioned by an administrator, or dynamically provisioned by Kubernetes using a StorageClass. A PersistentVolumeClaim (PVC) is a request for storage by a user that can be fulfilled by a PV. PersistentVolumes and PersistentVolumeClaims are independent from Pod lifecycles and preserve data through restarting, rescheduling, and even deleting Pods.

Objectives

  • Create PersistentVolumeClaims and PersistentVolumes
  • Create a kustomization.yaml with
    • a Secret generator
    • MySQL resource configs
    • WordPress resource configs
  • Apply the kustomization directory by kubectl apply -k ./
  • Clean up

Before you begin

You need to have a Kubernetes cluster, and the kubectl command-line tool must be configured to communicate with your cluster. It is recommended to run this tutorial on a cluster with at least two nodes that are not acting as control plane hosts. If you do not already have a cluster, you can create one by using minikube or you can use one of these Kubernetes playgrounds:

To check the version, enter kubectl version.

The example shown on this page works with kubectl 1.27 and above.

Download the following configuration files:

  1. mysql-deployment.yaml

  2. wordpress-deployment.yaml

Create PersistentVolumeClaims and PersistentVolumes

MySQL and Wordpress each require a PersistentVolume to store data. Their PersistentVolumeClaims will be created at the deployment step.

Many cluster environments have a default StorageClass installed. When a StorageClass is not specified in the PersistentVolumeClaim, the cluster's default StorageClass is used instead.

When a PersistentVolumeClaim is created, a PersistentVolume is dynamically provisioned based on the StorageClass configuration.

Create a kustomization.yaml

Add a Secret generator

A Secret is an object that stores a piece of sensitive data like a password or key. Since 1.14, kubectl supports the management of Kubernetes objects using a kustomization file. You can create a Secret by generators in kustomization.yaml.

Add a Secret generator in kustomization.yaml from the following command. You will need to replace YOUR_PASSWORD with the password you want to use.

cat <<EOF >./kustomization.yaml
secretGenerator:
- name: mysql-pass
  literals:
  - password=YOUR_PASSWORD
EOF

Add resource configs for MySQL and WordPress

The following manifest describes a single-instance MySQL Deployment. The MySQL container mounts the PersistentVolume at /var/lib/mysql. The MYSQL_ROOT_PASSWORD environment variable sets the database password from the Secret.

apiVersion: v1
kind: Service
metadata:
  name: wordpress-mysql
  labels:
    app: wordpress
spec:
  ports:
    - port: 3306
  selector:
    app: wordpress
    tier: mysql
  clusterIP: None
---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: mysql-pv-claim
  labels:
    app: wordpress
spec:
  accessModes:
    - ReadWriteOnce
  resources:
    requests:
      storage: 20Gi
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: wordpress-mysql
  labels:
    app: wordpress
spec:
  selector:
    matchLabels:
      app: wordpress
      tier: mysql
  strategy:
    type: Recreate
  template:
    metadata:
      labels:
        app: wordpress
        tier: mysql
    spec:
      containers:
      - image: mysql:8.0
        name: mysql
        env:
        - name: MYSQL_ROOT_PASSWORD
          valueFrom:
            secretKeyRef:
              name: mysql-pass
              key: password
        - name: MYSQL_DATABASE
          value: wordpress
        - name: MYSQL_USER
          value: wordpress
        - name: MYSQL_PASSWORD
          valueFrom:
            secretKeyRef:
              name: mysql-pass
              key: password
        ports:
        - containerPort: 3306
          name: mysql
        volumeMounts:
        - name: mysql-persistent-storage
          mountPath: /var/lib/mysql
      volumes:
      - name: mysql-persistent-storage
        persistentVolumeClaim:
          claimName: mysql-pv-claim

The following manifest describes a single-instance WordPress Deployment. The WordPress container mounts the PersistentVolume at /var/www/html for website data files. The WORDPRESS_DB_HOST environment variable sets the name of the MySQL Service defined above, and WordPress will access the database by Service. The WORDPRESS_DB_PASSWORD environment variable sets the database password from the Secret kustomize generated.

apiVersion: v1
kind: Service
metadata:
  name: wordpress
  labels:
    app: wordpress
spec:
  ports:
    - port: 80
  selector:
    app: wordpress
    tier: frontend
  type: LoadBalancer
---
apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: wp-pv-claim
  labels:
    app: wordpress
spec:
  accessModes:
    - ReadWriteOnce
  resources:
    requests:
      storage: 20Gi
---
apiVersion: apps/v1
kind: Deployment
metadata:
  name: wordpress
  labels:
    app: wordpress
spec:
  selector:
    matchLabels:
      app: wordpress
      tier: frontend
  strategy:
    type: Recreate
  template:
    metadata:
      labels:
        app: wordpress
        tier: frontend
    spec:
      containers:
      - image: wordpress:6.2.1-apache
        name: wordpress
        env:
        - name: WORDPRESS_DB_HOST
          value: wordpress-mysql
        - name: WORDPRESS_DB_PASSWORD
          valueFrom:
            secretKeyRef:
              name: mysql-pass
              key: password
        - name: WORDPRESS_DB_USER
          value: wordpress
        ports:
        - containerPort: 80
          name: wordpress
        volumeMounts:
        - name: wordpress-persistent-storage
          mountPath: /var/www/html
      volumes:
      - name: wordpress-persistent-storage
        persistentVolumeClaim:
          claimName: wp-pv-claim
  1. Download the MySQL deployment configuration file.

    curl -LO https://k8s.io/examples/application/wordpress/mysql-deployment.yaml
    
  2. Download the WordPress configuration file.

    curl -LO https://k8s.io/examples/application/wordpress/wordpress-deployment.yaml
    
  3. Add them to kustomization.yaml file.

    cat <<EOF >>./kustomization.yaml
    resources:
      - mysql-deployment.yaml
      - wordpress-deployment.yaml
    EOF
    

Apply and Verify

The kustomization.yaml contains all the resources for deploying a WordPress site and a MySQL database. You can apply the directory by

kubectl apply -k ./

Now you can verify that all objects exist.

  1. Verify that the Secret exists by running the following command:

    kubectl get secrets
    

    The response should be like this:

    NAME                    TYPE                                  DATA   AGE
    mysql-pass-c57bb4t7mf   Opaque                                1      9s
    
  2. Verify that a PersistentVolume got dynamically provisioned.

    kubectl get pvc
    

    The response should be like this:

    NAME             STATUS    VOLUME                                     CAPACITY   ACCESS MODES   STORAGECLASS       AGE
    mysql-pv-claim   Bound     pvc-8cbd7b2e-4044-11e9-b2bb-42010a800002   20Gi       RWO            standard           77s
    wp-pv-claim      Bound     pvc-8cd0df54-4044-11e9-b2bb-42010a800002   20Gi       RWO            standard           77s
    
  3. Verify that the Pod is running by running the following command:

    kubectl get pods
    

    The response should be like this:

    NAME                               READY     STATUS    RESTARTS   AGE
    wordpress-mysql-1894417608-x5dzt   1/1       Running   0          40s
    
  4. Verify that the Service is running by running the following command:

    kubectl get services wordpress
    

    The response should be like this:

    NAME        TYPE            CLUSTER-IP   EXTERNAL-IP   PORT(S)        AGE
    wordpress   LoadBalancer    10.0.0.89    <pending>     80:32406/TCP   4m
    
  5. Run the following command to get the IP Address for the WordPress Service:

    minikube service wordpress --url
    

    The response should be like this:

    http://1.2.3.4:32406
    
  6. Copy the IP address, and load the page in your browser to view your site.

    You should see the WordPress set up page similar to the following screenshot.

    wordpress-init

Cleaning up

  1. Run the following command to delete your Secret, Deployments, Services and PersistentVolumeClaims:

    kubectl delete -k ./
    

What's next

3 - Example: Deploying Cassandra with a StatefulSet

This tutorial shows you how to run Apache Cassandra on Kubernetes. Cassandra, a database, needs persistent storage to provide data durability (application state). In this example, a custom Cassandra seed provider lets the database discover new Cassandra instances as they join the Cassandra cluster.

StatefulSets make it easier to deploy stateful applications into your Kubernetes cluster. For more information on the features used in this tutorial, see StatefulSet.

Objectives

  • Create and validate a Cassandra headless Service.
  • Use a StatefulSet to create a Cassandra ring.
  • Validate the StatefulSet.
  • Modify the StatefulSet.
  • Delete the StatefulSet and its Pods.

Before you begin

You need to have a Kubernetes cluster, and the kubectl command-line tool must be configured to communicate with your cluster. It is recommended to run this tutorial on a cluster with at least two nodes that are not acting as control plane hosts. If you do not already have a cluster, you can create one by using minikube or you can use one of these Kubernetes playgrounds:

To complete this tutorial, you should already have a basic familiarity with Pods, Services, and StatefulSets.

Additional Minikube setup instructions

Creating a headless Service for Cassandra

In Kubernetes, a Service describes a set of Pods that perform the same task.

The following Service is used for DNS lookups between Cassandra Pods and clients within your cluster:

apiVersion: v1
kind: Service
metadata:
  labels:
    app: cassandra
  name: cassandra
spec:
  clusterIP: None
  ports:
  - port: 9042
  selector:
    app: cassandra

Create a Service to track all Cassandra StatefulSet members from the cassandra-service.yaml file:

kubectl apply -f https://k8s.io/examples/application/cassandra/cassandra-service.yaml

Validating (optional)

Get the Cassandra Service.

kubectl get svc cassandra

The response is

NAME        TYPE        CLUSTER-IP   EXTERNAL-IP   PORT(S)    AGE
cassandra   ClusterIP   None         <none>        9042/TCP   45s

If you don't see a Service named cassandra, that means creation failed. Read Debug Services for help troubleshooting common issues.

Using a StatefulSet to create a Cassandra ring

The StatefulSet manifest, included below, creates a Cassandra ring that consists of three Pods.

apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: cassandra
  labels:
    app: cassandra
spec:
  serviceName: cassandra
  replicas: 3
  selector:
    matchLabels:
      app: cassandra
  template:
    metadata:
      labels:
        app: cassandra
    spec:
      terminationGracePeriodSeconds: 1800
      containers:
      - name: cassandra
        image: gcr.io/google-samples/cassandra:v13
        imagePullPolicy: Always
        ports:
        - containerPort: 7000
          name: intra-node
        - containerPort: 7001
          name: tls-intra-node
        - containerPort: 7199
          name: jmx
        - containerPort: 9042
          name: cql
        resources:
          limits:
            cpu: "500m"
            memory: 1Gi
          requests:
            cpu: "500m"
            memory: 1Gi
        securityContext:
          capabilities:
            add:
              - IPC_LOCK
        lifecycle:
          preStop:
            exec:
              command: 
              - /bin/sh
              - -c
              - nodetool drain
        env:
          - name: MAX_HEAP_SIZE
            value: 512M
          - name: HEAP_NEWSIZE
            value: 100M
          - name: CASSANDRA_SEEDS
            value: "cassandra-0.cassandra.default.svc.cluster.local"
          - name: CASSANDRA_CLUSTER_NAME
            value: "K8Demo"
          - name: CASSANDRA_DC
            value: "DC1-K8Demo"
          - name: CASSANDRA_RACK
            value: "Rack1-K8Demo"
          - name: POD_IP
            valueFrom:
              fieldRef:
                fieldPath: status.podIP
        readinessProbe:
          exec:
            command:
            - /bin/bash
            - -c
            - /ready-probe.sh
          initialDelaySeconds: 15
          timeoutSeconds: 5
        # These volume mounts are persistent. They are like inline claims,
        # but not exactly because the names need to match exactly one of
        # the stateful pod volumes.
        volumeMounts:
        - name: cassandra-data
          mountPath: /cassandra_data
  # These are converted to volume claims by the controller
  # and mounted at the paths mentioned above.
  # do not use these in production until ssd GCEPersistentDisk or other ssd pd
  volumeClaimTemplates:
  - metadata:
      name: cassandra-data
    spec:
      accessModes: [ "ReadWriteOnce" ]
      storageClassName: fast
      resources:
        requests:
          storage: 1Gi
---
kind: StorageClass
apiVersion: storage.k8s.io/v1
metadata:
  name: fast
provisioner: k8s.io/minikube-hostpath
parameters:
  type: pd-ssd

Create the Cassandra StatefulSet from the cassandra-statefulset.yaml file:

# Use this if you are able to apply cassandra-statefulset.yaml unmodified
kubectl apply -f https://k8s.io/examples/application/cassandra/cassandra-statefulset.yaml

If you need to modify cassandra-statefulset.yaml to suit your cluster, download https://k8s.io/examples/application/cassandra/cassandra-statefulset.yaml and then apply that manifest, from the folder you saved the modified version into:

# Use this if you needed to modify cassandra-statefulset.yaml locally
kubectl apply -f cassandra-statefulset.yaml

Validating the Cassandra StatefulSet

  1. Get the Cassandra StatefulSet:

    kubectl get statefulset cassandra
    

    The response should be similar to:

    NAME        DESIRED   CURRENT   AGE
    cassandra   3         0         13s
    

    The StatefulSet resource deploys Pods sequentially.

  2. Get the Pods to see the ordered creation status:

    kubectl get pods -l="app=cassandra"
    

    The response should be similar to:

    NAME          READY     STATUS              RESTARTS   AGE
    cassandra-0   1/1       Running             0          1m
    cassandra-1   0/1       ContainerCreating   0          8s
    

    It can take several minutes for all three Pods to deploy. Once they are deployed, the same command returns output similar to:

    NAME          READY     STATUS    RESTARTS   AGE
    cassandra-0   1/1       Running   0          10m
    cassandra-1   1/1       Running   0          9m
    cassandra-2   1/1       Running   0          8m
    
  3. Run the Cassandra nodetool inside the first Pod, to display the status of the ring.

    kubectl exec -it cassandra-0 -- nodetool status
    

    The response should look something like:

    Datacenter: DC1-K8Demo
    ======================
    Status=Up/Down
    |/ State=Normal/Leaving/Joining/Moving
    --  Address     Load       Tokens       Owns (effective)  Host ID                               Rack
    UN  172.17.0.5  83.57 KiB  32           74.0%             e2dd09e6-d9d3-477e-96c5-45094c08db0f  Rack1-K8Demo
    UN  172.17.0.4  101.04 KiB  32           58.8%             f89d6835-3a42-4419-92b3-0e62cae1479c  Rack1-K8Demo
    UN  172.17.0.6  84.74 KiB  32           67.1%             a6a1e8c2-3dc5-4417-b1a0-26507af2aaad  Rack1-K8Demo
    

Modifying the Cassandra StatefulSet

Use kubectl edit to modify the size of a Cassandra StatefulSet.

  1. Run the following command:

    kubectl edit statefulset cassandra
    

    This command opens an editor in your terminal. The line you need to change is the replicas field. The following sample is an excerpt of the StatefulSet file:

    # Please edit the object below. Lines beginning with a '#' will be ignored,
    # and an empty file will abort the edit. If an error occurs while saving this file will be
    # reopened with the relevant failures.
    #
    apiVersion: apps/v1
    kind: StatefulSet
    metadata:
      creationTimestamp: 2016-08-13T18:40:58Z
      generation: 1
      labels:
      app: cassandra
      name: cassandra
      namespace: default
      resourceVersion: "323"
      uid: 7a219483-6185-11e6-a910-42010a8a0fc0
    spec:
      replicas: 3
    
  2. Change the number of replicas to 4, and then save the manifest.

    The StatefulSet now scales to run with 4 Pods.

  3. Get the Cassandra StatefulSet to verify your change:

    kubectl get statefulset cassandra
    

    The response should be similar to:

    NAME        DESIRED   CURRENT   AGE
    cassandra   4         4         36m
    

Cleaning up

Deleting or scaling a StatefulSet down does not delete the volumes associated with the StatefulSet. This setting is for your safety because your data is more valuable than automatically purging all related StatefulSet resources.

  1. Run the following commands (chained together into a single command) to delete everything in the Cassandra StatefulSet:

    grace=$(kubectl get pod cassandra-0 -o=jsonpath='{.spec.terminationGracePeriodSeconds}') \
      && kubectl delete statefulset -l app=cassandra \
      && echo "Sleeping ${grace} seconds" 1>&2 \
      && sleep $grace \
      && kubectl delete persistentvolumeclaim -l app=cassandra
    
  2. Run the following command to delete the Service you set up for Cassandra:

    kubectl delete service -l app=cassandra
    

Cassandra container environment variables

The Pods in this tutorial use the gcr.io/google-samples/cassandra:v13 image from Google's container registry. The Docker image above is based on debian-base and includes OpenJDK 8.

This image includes a standard Cassandra installation from the Apache Debian repo. By using environment variables you can change values that are inserted into cassandra.yaml.

Environment variableDefault value
CASSANDRA_CLUSTER_NAME'Test Cluster'
CASSANDRA_NUM_TOKENS32
CASSANDRA_RPC_ADDRESS0.0.0.0

What's next

4 - Running ZooKeeper, A Distributed System Coordinator

This tutorial demonstrates running Apache Zookeeper on Kubernetes using StatefulSets, PodDisruptionBudgets, and PodAntiAffinity.

Before you begin

Before starting this tutorial, you should be familiar with the following Kubernetes concepts:

You must have a cluster with at least four nodes, and each node requires at least 2 CPUs and 4 GiB of memory. In this tutorial you will cordon and drain the cluster's nodes. This means that the cluster will terminate and evict all Pods on its nodes, and the nodes will temporarily become unschedulable. You should use a dedicated cluster for this tutorial, or you should ensure that the disruption you cause will not interfere with other tenants.

This tutorial assumes that you have configured your cluster to dynamically provision PersistentVolumes. If your cluster is not configured to do so, you will have to manually provision three 20 GiB volumes before starting this tutorial.

Objectives

After this tutorial, you will know the following.

  • How to deploy a ZooKeeper ensemble using StatefulSet.
  • How to consistently configure the ensemble.
  • How to spread the deployment of ZooKeeper servers in the ensemble.
  • How to use PodDisruptionBudgets to ensure service availability during planned maintenance.

ZooKeeper

Apache ZooKeeper is a distributed, open-source coordination service for distributed applications. ZooKeeper allows you to read, write, and observe updates to data. Data are organized in a file system like hierarchy and replicated to all ZooKeeper servers in the ensemble (a set of ZooKeeper servers). All operations on data are atomic and sequentially consistent. ZooKeeper ensures this by using the Zab consensus protocol to replicate a state machine across all servers in the ensemble.

The ensemble uses the Zab protocol to elect a leader, and the ensemble cannot write data until that election is complete. Once complete, the ensemble uses Zab to ensure that it replicates all writes to a quorum before it acknowledges and makes them visible to clients. Without respect to weighted quorums, a quorum is a majority component of the ensemble containing the current leader. For instance, if the ensemble has three servers, a component that contains the leader and one other server constitutes a quorum. If the ensemble can not achieve a quorum, the ensemble cannot write data.

ZooKeeper servers keep their entire state machine in memory, and write every mutation to a durable WAL (Write Ahead Log) on storage media. When a server crashes, it can recover its previous state by replaying the WAL. To prevent the WAL from growing without bound, ZooKeeper servers will periodically snapshot them in memory state to storage media. These snapshots can be loaded directly into memory, and all WAL entries that preceded the snapshot may be discarded.

Creating a ZooKeeper ensemble

The manifest below contains a Headless Service, a Service, a PodDisruptionBudget, and a StatefulSet.

apiVersion: v1
kind: Service
metadata:
  name: zk-hs
  labels:
    app: zk
spec:
  ports:
  - port: 2888
    name: server
  - port: 3888
    name: leader-election
  clusterIP: None
  selector:
    app: zk
---
apiVersion: v1
kind: Service
metadata:
  name: zk-cs
  labels:
    app: zk
spec:
  ports:
  - port: 2181
    name: client
  selector:
    app: zk
---
apiVersion: policy/v1
kind: PodDisruptionBudget
metadata:
  name: zk-pdb
spec:
  selector:
    matchLabels:
      app: zk
  maxUnavailable: 1
---
apiVersion: apps/v1
kind: StatefulSet
metadata:
  name: zk
spec:
  selector:
    matchLabels:
      app: zk
  serviceName: zk-hs
  replicas: 3
  updateStrategy:
    type: RollingUpdate
  podManagementPolicy: OrderedReady
  template:
    metadata:
      labels:
        app: zk
    spec:
      affinity:
        podAntiAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
            - labelSelector:
                matchExpressions:
                  - key: "app"
                    operator: In
                    values:
                    - zk
              topologyKey: "kubernetes.io/hostname"
      containers:
      - name: kubernetes-zookeeper
        imagePullPolicy: Always
        image: "registry.k8s.io/kubernetes-zookeeper:1.0-3.4.10"
        resources:
          requests:
            memory: "1Gi"
            cpu: "0.5"
        ports:
        - containerPort: 2181
          name: client
        - containerPort: 2888
          name: server
        - containerPort: 3888
          name: leader-election
        command:
        - sh
        - -c
        - "start-zookeeper \
          --servers=3 \
          --data_dir=/var/lib/zookeeper/data \
          --data_log_dir=/var/lib/zookeeper/data/log \
          --conf_dir=/opt/zookeeper/conf \
          --client_port=2181 \
          --election_port=3888 \
          --server_port=2888 \
          --tick_time=2000 \
          --init_limit=10 \
          --sync_limit=5 \
          --heap=512M \
          --max_client_cnxns=60 \
          --snap_retain_count=3 \
          --purge_interval=12 \
          --max_session_timeout=40000 \
          --min_session_timeout=4000 \
          --log_level=INFO"
        readinessProbe:
          exec:
            command:
            - sh
            - -c
            - "zookeeper-ready 2181"
          initialDelaySeconds: 10
          timeoutSeconds: 5
        livenessProbe:
          exec:
            command:
            - sh
            - -c
            - "zookeeper-ready 2181"
          initialDelaySeconds: 10
          timeoutSeconds: 5
        volumeMounts:
        - name: datadir
          mountPath: /var/lib/zookeeper
      securityContext:
        runAsUser: 1000
        fsGroup: 1000
  volumeClaimTemplates:
  - metadata:
      name: datadir
    spec:
      accessModes: [ "ReadWriteOnce" ]
      resources:
        requests:
          storage: 10Gi

Open a terminal, and use the kubectl apply command to create the manifest.

kubectl apply -f https://k8s.io/examples/application/zookeeper/zookeeper.yaml

This creates the zk-hs Headless Service, the zk-cs Service, the zk-pdb PodDisruptionBudget, and the zk StatefulSet.

service/zk-hs created
service/zk-cs created
poddisruptionbudget.policy/zk-pdb created
statefulset.apps/zk created

Use kubectl get to watch the StatefulSet controller create the StatefulSet's Pods.

kubectl get pods -w -l app=zk

Once the zk-2 Pod is Running and Ready, use CTRL-C to terminate kubectl.

NAME      READY     STATUS    RESTARTS   AGE
zk-0      0/1       Pending   0          0s
zk-0      0/1       Pending   0         0s
zk-0      0/1       ContainerCreating   0         0s
zk-0      0/1       Running   0         19s
zk-0      1/1       Running   0         40s
zk-1      0/1       Pending   0         0s
zk-1      0/1       Pending   0         0s
zk-1      0/1       ContainerCreating   0         0s
zk-1      0/1       Running   0         18s
zk-1      1/1       Running   0         40s
zk-2      0/1       Pending   0         0s
zk-2      0/1       Pending   0         0s
zk-2      0/1       ContainerCreating   0         0s
zk-2      0/1       Running   0         19s
zk-2      1/1       Running   0         40s

The StatefulSet controller creates three Pods, and each Pod has a container with a ZooKeeper server.

Facilitating leader election

Because there is no terminating algorithm for electing a leader in an anonymous network, Zab requires explicit membership configuration to perform leader election. Each server in the ensemble needs to have a unique identifier, all servers need to know the global set of identifiers, and each identifier needs to be associated with a network address.

Use kubectl exec to get the hostnames of the Pods in the zk StatefulSet.

for i in 0 1 2; do kubectl exec zk-$i -- hostname; done

The StatefulSet controller provides each Pod with a unique hostname based on its ordinal index. The hostnames take the form of <statefulset name>-<ordinal index>. Because the replicas field of the zk StatefulSet is set to 3, the Set's controller creates three Pods with their hostnames set to zk-0, zk-1, and zk-2.

zk-0
zk-1
zk-2

The servers in a ZooKeeper ensemble use natural numbers as unique identifiers, and store each server's identifier in a file called myid in the server's data directory.

To examine the contents of the myid file for each server use the following command.

for i in 0 1 2; do echo "myid zk-$i";kubectl exec zk-$i -- cat /var/lib/zookeeper/data/myid; done

Because the identifiers are natural numbers and the ordinal indices are non-negative integers, you can generate an identifier by adding 1 to the ordinal.

myid zk-0
1
myid zk-1
2
myid zk-2
3

To get the Fully Qualified Domain Name (FQDN) of each Pod in the zk StatefulSet use the following command.

for i in 0 1 2; do kubectl exec zk-$i -- hostname -f; done

The zk-hs Service creates a domain for all of the Pods, zk-hs.default.svc.cluster.local.

zk-0.zk-hs.default.svc.cluster.local
zk-1.zk-hs.default.svc.cluster.local
zk-2.zk-hs.default.svc.cluster.local

The A records in Kubernetes DNS resolve the FQDNs to the Pods' IP addresses. If Kubernetes reschedules the Pods, it will update the A records with the Pods' new IP addresses, but the A records names will not change.

ZooKeeper stores its application configuration in a file named zoo.cfg. Use kubectl exec to view the contents of the zoo.cfg file in the zk-0 Pod.

kubectl exec zk-0 -- cat /opt/zookeeper/conf/zoo.cfg

In the server.1, server.2, and server.3 properties at the bottom of the file, the 1, 2, and 3 correspond to the identifiers in the ZooKeeper servers' myid files. They are set to the FQDNs for the Pods in the zk StatefulSet.

clientPort=2181
dataDir=/var/lib/zookeeper/data
dataLogDir=/var/lib/zookeeper/log
tickTime=2000
initLimit=10
syncLimit=2000
maxClientCnxns=60
minSessionTimeout= 4000
maxSessionTimeout= 40000
autopurge.snapRetainCount=3
autopurge.purgeInterval=0
server.1=zk-0.zk-hs.default.svc.cluster.local:2888:3888
server.2=zk-1.zk-hs.default.svc.cluster.local:2888:3888
server.3=zk-2.zk-hs.default.svc.cluster.local:2888:3888

Achieving consensus

Consensus protocols require that the identifiers of each participant be unique. No two participants in the Zab protocol should claim the same unique identifier. This is necessary to allow the processes in the system to agree on which processes have committed which data. If two Pods are launched with the same ordinal, two ZooKeeper servers would both identify themselves as the same server.

kubectl get pods -w -l app=zk
NAME      READY     STATUS    RESTARTS   AGE
zk-0      0/1       Pending   0          0s
zk-0      0/1       Pending   0         0s
zk-0      0/1       ContainerCreating   0         0s
zk-0      0/1       Running   0         19s
zk-0      1/1       Running   0         40s
zk-1      0/1       Pending   0         0s
zk-1      0/1       Pending   0         0s
zk-1      0/1       ContainerCreating   0         0s
zk-1      0/1       Running   0         18s
zk-1      1/1       Running   0         40s
zk-2      0/1       Pending   0         0s
zk-2      0/1       Pending   0         0s
zk-2      0/1       ContainerCreating   0         0s
zk-2      0/1       Running   0         19s
zk-2      1/1       Running   0         40s

The A records for each Pod are entered when the Pod becomes Ready. Therefore, the FQDNs of the ZooKeeper servers will resolve to a single endpoint, and that endpoint will be the unique ZooKeeper server claiming the identity configured in its myid file.

zk-0.zk-hs.default.svc.cluster.local
zk-1.zk-hs.default.svc.cluster.local
zk-2.zk-hs.default.svc.cluster.local

This ensures that the servers properties in the ZooKeepers' zoo.cfg files represents a correctly configured ensemble.

server.1=zk-0.zk-hs.default.svc.cluster.local:2888:3888
server.2=zk-1.zk-hs.default.svc.cluster.local:2888:3888
server.3=zk-2.zk-hs.default.svc.cluster.local:2888:3888

When the servers use the Zab protocol to attempt to commit a value, they will either achieve consensus and commit the value (if leader election has succeeded and at least two of the Pods are Running and Ready), or they will fail to do so (if either of the conditions are not met). No state will arise where one server acknowledges a write on behalf of another.

Sanity testing the ensemble

The most basic sanity test is to write data to one ZooKeeper server and to read the data from another.

The command below executes the zkCli.sh script to write world to the path /hello on the zk-0 Pod in the ensemble.

kubectl exec zk-0 -- zkCli.sh create /hello world
WATCHER::

WatchedEvent state:SyncConnected type:None path:null
Created /hello

To get the data from the zk-1 Pod use the following command.

kubectl exec zk-1 -- zkCli.sh get /hello

The data that you created on zk-0 is available on all the servers in the ensemble.

WATCHER::

WatchedEvent state:SyncConnected type:None path:null
world
cZxid = 0x100000002
ctime = Thu Dec 08 15:13:30 UTC 2016
mZxid = 0x100000002
mtime = Thu Dec 08 15:13:30 UTC 2016
pZxid = 0x100000002
cversion = 0
dataVersion = 0
aclVersion = 0
ephemeralOwner = 0x0
dataLength = 5
numChildren = 0

Providing durable storage

As mentioned in the ZooKeeper Basics section, ZooKeeper commits all entries to a durable WAL, and periodically writes snapshots in memory state, to storage media. Using WALs to provide durability is a common technique for applications that use consensus protocols to achieve a replicated state machine.

Use the kubectl delete command to delete the zk StatefulSet.

kubectl delete statefulset zk
statefulset.apps "zk" deleted

Watch the termination of the Pods in the StatefulSet.

kubectl get pods -w -l app=zk

When zk-0 if fully terminated, use CTRL-C to terminate kubectl.

zk-2      1/1       Terminating   0         9m
zk-0      1/1       Terminating   0         11m
zk-1      1/1       Terminating   0         10m
zk-2      0/1       Terminating   0         9m
zk-2      0/1       Terminating   0         9m
zk-2      0/1       Terminating   0         9m
zk-1      0/1       Terminating   0         10m
zk-1      0/1       Terminating   0         10m
zk-1      0/1       Terminating   0         10m
zk-0      0/1       Terminating   0         11m
zk-0      0/1       Terminating   0         11m
zk-0      0/1       Terminating   0         11m

Reapply the manifest in zookeeper.yaml.

kubectl apply -f https://k8s.io/examples/application/zookeeper/zookeeper.yaml

This creates the zk StatefulSet object, but the other API objects in the manifest are not modified because they already exist.

Watch the StatefulSet controller recreate the StatefulSet's Pods.

kubectl get pods -w -l app=zk

Once the zk-2 Pod is Running and Ready, use CTRL-C to terminate kubectl.

NAME      READY     STATUS    RESTARTS   AGE
zk-0      0/1       Pending   0          0s
zk-0      0/1       Pending   0         0s
zk-0      0/1       ContainerCreating   0         0s
zk-0      0/1       Running   0         19s
zk-0      1/1       Running   0         40s
zk-1      0/1       Pending   0         0s
zk-1      0/1       Pending   0         0s
zk-1      0/1       ContainerCreating   0         0s
zk-1      0/1       Running   0         18s
zk-1      1/1       Running   0         40s
zk-2      0/1       Pending   0         0s
zk-2      0/1       Pending   0         0s
zk-2      0/1       ContainerCreating   0         0s
zk-2      0/1       Running   0         19s
zk-2      1/1       Running   0         40s

Use the command below to get the value you entered during the sanity test, from the zk-2 Pod.

kubectl exec zk-2 zkCli.sh get /hello

Even though you terminated and recreated all of the Pods in the zk StatefulSet, the ensemble still serves the original value.

WATCHER::

WatchedEvent state:SyncConnected type:None path:null
world
cZxid = 0x100000002
ctime = Thu Dec 08 15:13:30 UTC 2016
mZxid = 0x100000002
mtime = Thu Dec 08 15:13:30 UTC 2016
pZxid = 0x100000002
cversion = 0
dataVersion = 0
aclVersion = 0
ephemeralOwner = 0x0
dataLength = 5
numChildren = 0

The volumeClaimTemplates field of the zk StatefulSet's spec specifies a PersistentVolume provisioned for each Pod.

volumeClaimTemplates:
  - metadata:
      name: datadir
      annotations:
        volume.alpha.kubernetes.io/storage-class: anything
    spec:
      accessModes: [ "ReadWriteOnce" ]
      resources:
        requests:
          storage: 20Gi

The StatefulSet controller generates a PersistentVolumeClaim for each Pod in the StatefulSet.

Use the following command to get the StatefulSet's PersistentVolumeClaims.

kubectl get pvc -l app=zk

When the StatefulSet recreated its Pods, it remounts the Pods' PersistentVolumes.

NAME           STATUS    VOLUME                                     CAPACITY   ACCESSMODES   AGE
datadir-zk-0   Bound     pvc-bed742cd-bcb1-11e6-994f-42010a800002   20Gi       RWO           1h
datadir-zk-1   Bound     pvc-bedd27d2-bcb1-11e6-994f-42010a800002   20Gi       RWO           1h
datadir-zk-2   Bound     pvc-bee0817e-bcb1-11e6-994f-42010a800002   20Gi       RWO           1h

The volumeMounts section of the StatefulSet's container template mounts the PersistentVolumes in the ZooKeeper servers' data directories.

volumeMounts:
- name: datadir
  mountPath: /var/lib/zookeeper

When a Pod in the zk StatefulSet is (re)scheduled, it will always have the same PersistentVolume mounted to the ZooKeeper server's data directory. Even when the Pods are rescheduled, all the writes made to the ZooKeeper servers' WALs, and all their snapshots, remain durable.

Ensuring consistent configuration

As noted in the Facilitating Leader Election and Achieving Consensus sections, the servers in a ZooKeeper ensemble require consistent configuration to elect a leader and form a quorum. They also require consistent configuration of the Zab protocol in order for the protocol to work correctly over a network. In our example we achieve consistent configuration by embedding the configuration directly into the manifest.

Get the zk StatefulSet.

kubectl get sts zk -o yaml
…
command:
      - sh
      - -c
      - "start-zookeeper \
        --servers=3 \
        --data_dir=/var/lib/zookeeper/data \
        --data_log_dir=/var/lib/zookeeper/data/log \
        --conf_dir=/opt/zookeeper/conf \
        --client_port=2181 \
        --election_port=3888 \
        --server_port=2888 \
        --tick_time=2000 \
        --init_limit=10 \
        --sync_limit=5 \
        --heap=512M \
        --max_client_cnxns=60 \
        --snap_retain_count=3 \
        --purge_interval=12 \
        --max_session_timeout=40000 \
        --min_session_timeout=4000 \
        --log_level=INFO"
…

The command used to start the ZooKeeper servers passed the configuration as command line parameter. You can also use environment variables to pass configuration to the ensemble.

Configuring logging

One of the files generated by the zkGenConfig.sh script controls ZooKeeper's logging. ZooKeeper uses Log4j, and, by default, it uses a time and size based rolling file appender for its logging configuration.

Use the command below to get the logging configuration from one of Pods in the zk StatefulSet.

kubectl exec zk-0 cat /usr/etc/zookeeper/log4j.properties

The logging configuration below will cause the ZooKeeper process to write all of its logs to the standard output file stream.

zookeeper.root.logger=CONSOLE
zookeeper.console.threshold=INFO
log4j.rootLogger=${zookeeper.root.logger}
log4j.appender.CONSOLE=org.apache.log4j.ConsoleAppender
log4j.appender.CONSOLE.Threshold=${zookeeper.console.threshold}
log4j.appender.CONSOLE.layout=org.apache.log4j.PatternLayout
log4j.appender.CONSOLE.layout.ConversionPattern=%d{ISO8601} [myid:%X{myid}] - %-5p [%t:%C{1}@%L] - %m%n

This is the simplest possible way to safely log inside the container. Because the applications write logs to standard out, Kubernetes will handle log rotation for you. Kubernetes also implements a sane retention policy that ensures application logs written to standard out and standard error do not exhaust local storage media.

Use kubectl logs to retrieve the last 20 log lines from one of the Pods.

kubectl logs zk-0 --tail 20

You can view application logs written to standard out or standard error using kubectl logs and from the Kubernetes Dashboard.

2016-12-06 19:34:16,236 [myid:1] - INFO  [NIOServerCxn.Factory:0.0.0.0/0.0.0.0:2181:NIOServerCnxn@827] - Processing ruok command from /127.0.0.1:52740
2016-12-06 19:34:16,237 [myid:1] - INFO  [Thread-1136:NIOServerCnxn@1008] - Closed socket connection for client /127.0.0.1:52740 (no session established for client)
2016-12-06 19:34:26,155 [myid:1] - INFO  [NIOServerCxn.Factory:0.0.0.0/0.0.0.0:2181:NIOServerCnxnFactory@192] - Accepted socket connection from /127.0.0.1:52749
2016-12-06 19:34:26,155 [myid:1] - INFO  [NIOServerCxn.Factory:0.0.0.0/0.0.0.0:2181:NIOServerCnxn@827] - Processing ruok command from /127.0.0.1:52749
2016-12-06 19:34:26,156 [myid:1] - INFO  [Thread-1137:NIOServerCnxn@1008] - Closed socket connection for client /127.0.0.1:52749 (no session established for client)
2016-12-06 19:34:26,222 [myid:1] - INFO  [NIOServerCxn.Factory:0.0.0.0/0.0.0.0:2181:NIOServerCnxnFactory@192] - Accepted socket connection from /127.0.0.1:52750
2016-12-06 19:34:26,222 [myid:1] - INFO  [NIOServerCxn.Factory:0.0.0.0/0.0.0.0:2181:NIOServerCnxn@827] - Processing ruok command from /127.0.0.1:52750
2016-12-06 19:34:26,226 [myid:1] - INFO  [Thread-1138:NIOServerCnxn@1008] - Closed socket connection for client /127.0.0.1:52750 (no session established for client)
2016-12-06 19:34:36,151 [myid:1] - INFO  [NIOServerCxn.Factory:0.0.0.0/0.0.0.0:2181:NIOServerCnxnFactory@192] - Accepted socket connection from /127.0.0.1:52760
2016-12-06 19:34:36,152 [myid:1] - INFO  [NIOServerCxn.Factory:0.0.0.0/0.0.0.0:2181:NIOServerCnxn@827] - Processing ruok command from /127.0.0.1:52760
2016-12-06 19:34:36,152 [myid:1] - INFO  [Thread-1139:NIOServerCnxn@1008] - Closed socket connection for client /127.0.0.1:52760 (no session established for client)
2016-12-06 19:34:36,230 [myid:1] - INFO  [NIOServerCxn.Factory:0.0.0.0/0.0.0.0:2181:NIOServerCnxnFactory@192] - Accepted socket connection from /127.0.0.1:52761
2016-12-06 19:34:36,231 [myid:1] - INFO  [NIOServerCxn.Factory:0.0.0.0/0.0.0.0:2181:NIOServerCnxn@827] - Processing ruok command from /127.0.0.1:52761
2016-12-06 19:34:36,231 [myid:1] - INFO  [Thread-1140:NIOServerCnxn@1008] - Closed socket connection for client /127.0.0.1:52761 (no session established for client)
2016-12-06 19:34:46,149 [myid:1] - INFO  [NIOServerCxn.Factory:0.0.0.0/0.0.0.0:2181:NIOServerCnxnFactory@192] - Accepted socket connection from /127.0.0.1:52767
2016-12-06 19:34:46,149 [myid:1] - INFO  [NIOServerCxn.Factory:0.0.0.0/0.0.0.0:2181:NIOServerCnxn@827] - Processing ruok command from /127.0.0.1:52767
2016-12-06 19:34:46,149 [myid:1] - INFO  [Thread-1141:NIOServerCnxn@1008] - Closed socket connection for client /127.0.0.1:52767 (no session established for client)
2016-12-06 19:34:46,230 [myid:1] - INFO  [NIOServerCxn.Factory:0.0.0.0/0.0.0.0:2181:NIOServerCnxnFactory@192] - Accepted socket connection from /127.0.0.1:52768
2016-12-06 19:34:46,230 [myid:1] - INFO  [NIOServerCxn.Factory:0.0.0.0/0.0.0.0:2181:NIOServerCnxn@827] - Processing ruok command from /127.0.0.1:52768
2016-12-06 19:34:46,230 [myid:1] - INFO  [Thread-1142:NIOServerCnxn@1008] - Closed socket connection for client /127.0.0.1:52768 (no session established for client)

Kubernetes integrates with many logging solutions. You can choose a logging solution that best fits your cluster and applications. For cluster-level logging and aggregation, consider deploying a sidecar container to rotate and ship your logs.

Configuring a non-privileged user

The best practices to allow an application to run as a privileged user inside of a container are a matter of debate. If your organization requires that applications run as a non-privileged user you can use a SecurityContext to control the user that the entry point runs as.

The zk StatefulSet's Pod template contains a SecurityContext.

securityContext:
  runAsUser: 1000
  fsGroup: 1000

In the Pods' containers, UID 1000 corresponds to the zookeeper user and GID 1000 corresponds to the zookeeper group.

Get the ZooKeeper process information from the zk-0 Pod.

kubectl exec zk-0 -- ps -elf

As the runAsUser field of the securityContext object is set to 1000, instead of running as root, the ZooKeeper process runs as the zookeeper user.

F S UID        PID  PPID  C PRI  NI ADDR SZ WCHAN  STIME TTY          TIME CMD
4 S zookeep+     1     0  0  80   0 -  1127 -      20:46 ?        00:00:00 sh -c zkGenConfig.sh && zkServer.sh start-foreground
0 S zookeep+    27     1  0  80   0 - 1155556 -    20:46 ?        00:00:19 /usr/lib/jvm/java-8-openjdk-amd64/bin/java -Dzookeeper.log.dir=/var/log/zookeeper -Dzookeeper.root.logger=INFO,CONSOLE -cp /usr/bin/../build/classes:/usr/bin/../build/lib/*.jar:/usr/bin/../share/zookeeper/zookeeper-3.4.9.jar:/usr/bin/../share/zookeeper/slf4j-log4j12-1.6.1.jar:/usr/bin/../share/zookeeper/slf4j-api-1.6.1.jar:/usr/bin/../share/zookeeper/netty-3.10.5.Final.jar:/usr/bin/../share/zookeeper/log4j-1.2.16.jar:/usr/bin/../share/zookeeper/jline-0.9.94.jar:/usr/bin/../src/java/lib/*.jar:/usr/bin/../etc/zookeeper: -Xmx2G -Xms2G -Dcom.sun.management.jmxremote -Dcom.sun.management.jmxremote.local.only=false org.apache.zookeeper.server.quorum.QuorumPeerMain /usr/bin/../etc/zookeeper/zoo.cfg

By default, when the Pod's PersistentVolumes is mounted to the ZooKeeper server's data directory, it is only accessible by the root user. This configuration prevents the ZooKeeper process from writing to its WAL and storing its snapshots.

Use the command below to get the file permissions of the ZooKeeper data directory on the zk-0 Pod.

kubectl exec -ti zk-0 -- ls -ld /var/lib/zookeeper/data

Because the fsGroup field of the securityContext object is set to 1000, the ownership of the Pods' PersistentVolumes is set to the zookeeper group, and the ZooKeeper process is able to read and write its data.

drwxr-sr-x 3 zookeeper zookeeper 4096 Dec  5 20:45 /var/lib/zookeeper/data

Managing the ZooKeeper process

The ZooKeeper documentation mentions that "You will want to have a supervisory process that manages each of your ZooKeeper server processes (JVM)." Utilizing a watchdog (supervisory process) to restart failed processes in a distributed system is a common pattern. When deploying an application in Kubernetes, rather than using an external utility as a supervisory process, you should use Kubernetes as the watchdog for your application.

Updating the ensemble

The zk StatefulSet is configured to use the RollingUpdate update strategy.

You can use kubectl patch to update the number of cpus allocated to the servers.

kubectl patch sts zk --type='json' -p='[{"op": "replace", "path": "/spec/template/spec/containers/0/resources/requests/cpu", "value":"0.3"}]'
statefulset.apps/zk patched

Use kubectl rollout status to watch the status of the update.

kubectl rollout status sts/zk
waiting for statefulset rolling update to complete 0 pods at revision zk-5db4499664...
Waiting for 1 pods to be ready...
Waiting for 1 pods to be ready...
waiting for statefulset rolling update to complete 1 pods at revision zk-5db4499664...
Waiting for 1 pods to be ready...
Waiting for 1 pods to be ready...
waiting for statefulset rolling update to complete 2 pods at revision zk-5db4499664...
Waiting for 1 pods to be ready...
Waiting for 1 pods to be ready...
statefulset rolling update complete 3 pods at revision zk-5db4499664...

This terminates the Pods, one at a time, in reverse ordinal order, and recreates them with the new configuration. This ensures that quorum is maintained during a rolling update.

Use the kubectl rollout history command to view a history or previous configurations.

kubectl rollout history sts/zk

The output is similar to this:

statefulsets "zk"
REVISION
1
2

Use the kubectl rollout undo command to roll back the modification.

kubectl rollout undo sts/zk

The output is similar to this:

statefulset.apps/zk rolled back

Handling process failure

Restart Policies control how Kubernetes handles process failures for the entry point of the container in a Pod. For Pods in a StatefulSet, the only appropriate RestartPolicy is Always, and this is the default value. For stateful applications you should never override the default policy.

Use the following command to examine the process tree for the ZooKeeper server running in the zk-0 Pod.

kubectl exec zk-0 -- ps -ef

The command used as the container's entry point has PID 1, and the ZooKeeper process, a child of the entry point, has PID 27.

UID        PID  PPID  C STIME TTY          TIME CMD
zookeep+     1     0  0 15:03 ?        00:00:00 sh -c zkGenConfig.sh && zkServer.sh start-foreground
zookeep+    27     1  0 15:03 ?        00:00:03 /usr/lib/jvm/java-8-openjdk-amd64/bin/java -Dzookeeper.log.dir=/var/log/zookeeper -Dzookeeper.root.logger=INFO,CONSOLE -cp /usr/bin/../build/classes:/usr/bin/../build/lib/*.jar:/usr/bin/../share/zookeeper/zookeeper-3.4.9.jar:/usr/bin/../share/zookeeper/slf4j-log4j12-1.6.1.jar:/usr/bin/../share/zookeeper/slf4j-api-1.6.1.jar:/usr/bin/../share/zookeeper/netty-3.10.5.Final.jar:/usr/bin/../share/zookeeper/log4j-1.2.16.jar:/usr/bin/../share/zookeeper/jline-0.9.94.jar:/usr/bin/../src/java/lib/*.jar:/usr/bin/../etc/zookeeper: -Xmx2G -Xms2G -Dcom.sun.management.jmxremote -Dcom.sun.management.jmxremote.local.only=false org.apache.zookeeper.server.quorum.QuorumPeerMain /usr/bin/../etc/zookeeper/zoo.cfg

In another terminal watch the Pods in the zk StatefulSet with the following command.

kubectl get pod -w -l app=zk

In another terminal, terminate the ZooKeeper process in Pod zk-0 with the following command.

kubectl exec zk-0 -- pkill java

The termination of the ZooKeeper process caused its parent process to terminate. Because the RestartPolicy of the container is Always, it restarted the parent process.

NAME      READY     STATUS    RESTARTS   AGE
zk-0      1/1       Running   0          21m
zk-1      1/1       Running   0          20m
zk-2      1/1       Running   0          19m
NAME      READY     STATUS    RESTARTS   AGE
zk-0      0/1       Error     0          29m
zk-0      0/1       Running   1         29m
zk-0      1/1       Running   1         29m

If your application uses a script (such as zkServer.sh) to launch the process that implements the application's business logic, the script must terminate with the child process. This ensures that Kubernetes will restart the application's container when the process implementing the application's business logic fails.

Testing for liveness

Configuring your application to restart failed processes is not enough to keep a distributed system healthy. There are scenarios where a system's processes can be both alive and unresponsive, or otherwise unhealthy. You should use liveness probes to notify Kubernetes that your application's processes are unhealthy and it should restart them.

The Pod template for the zk StatefulSet specifies a liveness probe.

  livenessProbe:
    exec:
      command:
      - sh
      - -c
      - "zookeeper-ready 2181"
    initialDelaySeconds: 15
    timeoutSeconds: 5

The probe calls a bash script that uses the ZooKeeper ruok four letter word to test the server's health.

OK=$(echo ruok | nc 127.0.0.1 $1)
if [ "$OK" == "imok" ]; then
    exit 0
else
    exit 1
fi

In one terminal window, use the following command to watch the Pods in the zk StatefulSet.

kubectl get pod -w -l app=zk

In another window, using the following command to delete the zookeeper-ready script from the file system of Pod zk-0.

kubectl exec zk-0 -- rm /opt/zookeeper/bin/zookeeper-ready

When the liveness probe for the ZooKeeper process fails, Kubernetes will automatically restart the process for you, ensuring that unhealthy processes in the ensemble are restarted.

kubectl get pod -w -l app=zk
NAME      READY     STATUS    RESTARTS   AGE
zk-0      1/1       Running   0          1h
zk-1      1/1       Running   0          1h
zk-2      1/1       Running   0          1h
NAME      READY     STATUS    RESTARTS   AGE
zk-0      0/1       Running   0          1h
zk-0      0/1       Running   1         1h
zk-0      1/1       Running   1         1h

Testing for readiness

Readiness is not the same as liveness. If a process is alive, it is scheduled and healthy. If a process is ready, it is able to process input. Liveness is a necessary, but not sufficient, condition for readiness. There are cases, particularly during initialization and termination, when a process can be alive but not ready.

If you specify a readiness probe, Kubernetes will ensure that your application's processes will not receive network traffic until their readiness checks pass.

For a ZooKeeper server, liveness implies readiness. Therefore, the readiness probe from the zookeeper.yaml manifest is identical to the liveness probe.

  readinessProbe:
    exec:
      command:
      - sh
      - -c
      - "zookeeper-ready 2181"
    initialDelaySeconds: 15
    timeoutSeconds: 5

Even though the liveness and readiness probes are identical, it is important to specify both. This ensures that only healthy servers in the ZooKeeper ensemble receive network traffic.

Tolerating Node failure

ZooKeeper needs a quorum of servers to successfully commit mutations to data. For a three server ensemble, two servers must be healthy for writes to succeed. In quorum based systems, members are deployed across failure domains to ensure availability. To avoid an outage, due to the loss of an individual machine, best practices preclude co-locating multiple instances of the application on the same machine.

By default, Kubernetes may co-locate Pods in a StatefulSet on the same node. For the three server ensemble you created, if two servers are on the same node, and that node fails, the clients of your ZooKeeper service will experience an outage until at least one of the Pods can be rescheduled.

You should always provision additional capacity to allow the processes of critical systems to be rescheduled in the event of node failures. If you do so, then the outage will only last until the Kubernetes scheduler reschedules one of the ZooKeeper servers. However, if you want your service to tolerate node failures with no downtime, you should set podAntiAffinity.

Use the command below to get the nodes for Pods in the zk StatefulSet.

for i in 0 1 2; do kubectl get pod zk-$i --template {{.spec.nodeName}}; echo ""; done

All of the Pods in the zk StatefulSet are deployed on different nodes.

kubernetes-node-cxpk
kubernetes-node-a5aq
kubernetes-node-2g2d

This is because the Pods in the zk StatefulSet have a PodAntiAffinity specified.

affinity:
  podAntiAffinity:
    requiredDuringSchedulingIgnoredDuringExecution:
      - labelSelector:
          matchExpressions:
            - key: "app"
              operator: In
              values:
                - zk
        topologyKey: "kubernetes.io/hostname"

The requiredDuringSchedulingIgnoredDuringExecution field tells the Kubernetes Scheduler that it should never co-locate two Pods which have app label as zk in the domain defined by the topologyKey. The topologyKey kubernetes.io/hostname indicates that the domain is an individual node. Using different rules, labels, and selectors, you can extend this technique to spread your ensemble across physical, network, and power failure domains.

Surviving maintenance

In this section you will cordon and drain nodes. If you are using this tutorial on a shared cluster, be sure that this will not adversely affect other tenants.

The previous section showed you how to spread your Pods across nodes to survive unplanned node failures, but you also need to plan for temporary node failures that occur due to planned maintenance.

Use this command to get the nodes in your cluster.

kubectl get nodes

This tutorial assumes a cluster with at least four nodes. If the cluster has more than four, use kubectl cordon to cordon all but four nodes. Constraining to four nodes will ensure Kubernetes encounters affinity and PodDisruptionBudget constraints when scheduling zookeeper Pods in the following maintenance simulation.

kubectl cordon <node-name>

Use this command to get the zk-pdb PodDisruptionBudget.

kubectl get pdb zk-pdb

The max-unavailable field indicates to Kubernetes that at most one Pod from zk StatefulSet can be unavailable at any time.

NAME      MIN-AVAILABLE   MAX-UNAVAILABLE   ALLOWED-DISRUPTIONS   AGE
zk-pdb    N/A             1                 1

In one terminal, use this command to watch the Pods in the zk StatefulSet.

kubectl get pods -w -l app=zk

In another terminal, use this command to get the nodes that the Pods are currently scheduled on.

for i in 0 1 2; do kubectl get pod zk-$i --template {{.spec.nodeName}}; echo ""; done

The output is similar to this:

kubernetes-node-pb41
kubernetes-node-ixsl
kubernetes-node-i4c4

Use kubectl drain to cordon and drain the node on which the zk-0 Pod is scheduled.

kubectl drain $(kubectl get pod zk-0 --template {{.spec.nodeName}}) --ignore-daemonsets --force --delete-emptydir-data

The output is similar to this:

node "kubernetes-node-pb41" cordoned

WARNING: Deleting pods not managed by ReplicationController, ReplicaSet, Job, or DaemonSet: fluentd-cloud-logging-kubernetes-node-pb41, kube-proxy-kubernetes-node-pb41; Ignoring DaemonSet-managed pods: node-problem-detector-v0.1-o5elz
pod "zk-0" deleted
node "kubernetes-node-pb41" drained

As there are four nodes in your cluster, kubectl drain, succeeds and the zk-0 is rescheduled to another node.

NAME      READY     STATUS    RESTARTS   AGE
zk-0      1/1       Running   2          1h
zk-1      1/1       Running   0          1h
zk-2      1/1       Running   0          1h
NAME      READY     STATUS        RESTARTS   AGE
zk-0      1/1       Terminating   2          2h
zk-0      0/1       Terminating   2         2h
zk-0      0/1       Terminating   2         2h
zk-0      0/1       Terminating   2         2h
zk-0      0/1       Pending   0         0s
zk-0      0/1       Pending   0         0s
zk-0      0/1       ContainerCreating   0         0s
zk-0      0/1       Running   0         51s
zk-0      1/1       Running   0         1m

Keep watching the StatefulSet's Pods in the first terminal and drain the node on which zk-1 is scheduled.

kubectl drain $(kubectl get pod zk-1 --template {{.spec.nodeName}}) --ignore-daemonsets --force --delete-emptydir-data

The output is similar to this:

"kubernetes-node-ixsl" cordoned
WARNING: Deleting pods not managed by ReplicationController, ReplicaSet, Job, or DaemonSet: fluentd-cloud-logging-kubernetes-node-ixsl, kube-proxy-kubernetes-node-ixsl; Ignoring DaemonSet-managed pods: node-problem-detector-v0.1-voc74
pod "zk-1" deleted
node "kubernetes-node-ixsl" drained

The zk-1 Pod cannot be scheduled because the zk StatefulSet contains a PodAntiAffinity rule preventing co-location of the Pods, and as only two nodes are schedulable, the Pod will remain in a Pending state.

kubectl get pods -w -l app=zk

The output is similar to this:

NAME      READY     STATUS    RESTARTS   AGE
zk-0      1/1       Running   2          1h
zk-1      1/1       Running   0          1h
zk-2      1/1       Running   0          1h
NAME      READY     STATUS        RESTARTS   AGE
zk-0      1/1       Terminating   2          2h
zk-0      0/1       Terminating   2         2h
zk-0      0/1       Terminating   2         2h
zk-0      0/1       Terminating   2         2h
zk-0      0/1       Pending   0         0s
zk-0      0/1       Pending   0         0s
zk-0      0/1       ContainerCreating   0         0s
zk-0      0/1       Running   0         51s
zk-0      1/1       Running   0         1m
zk-1      1/1       Terminating   0         2h
zk-1      0/1       Terminating   0         2h
zk-1      0/1       Terminating   0         2h
zk-1      0/1       Terminating   0         2h
zk-1      0/1       Pending   0         0s
zk-1      0/1       Pending   0         0s

Continue to watch the Pods of the StatefulSet, and drain the node on which zk-2 is scheduled.

kubectl drain $(kubectl get pod zk-2 --template {{.spec.nodeName}}) --ignore-daemonsets --force --delete-emptydir-data

The output is similar to this:

node "kubernetes-node-i4c4" cordoned

WARNING: Deleting pods not managed by ReplicationController, ReplicaSet, Job, or DaemonSet: fluentd-cloud-logging-kubernetes-node-i4c4, kube-proxy-kubernetes-node-i4c4; Ignoring DaemonSet-managed pods: node-problem-detector-v0.1-dyrog
WARNING: Ignoring DaemonSet-managed pods: node-problem-detector-v0.1-dyrog; Deleting pods not managed by ReplicationController, ReplicaSet, Job, or DaemonSet: fluentd-cloud-logging-kubernetes-node-i4c4, kube-proxy-kubernetes-node-i4c4
There are pending pods when an error occurred: Cannot evict pod as it would violate the pod's disruption budget.
pod/zk-2

Use CTRL-C to terminate kubectl.

You cannot drain the third node because evicting zk-2 would violate zk-budget. However, the node will remain cordoned.

Use zkCli.sh to retrieve the value you entered during the sanity test from zk-0.

kubectl exec zk-0 zkCli.sh get /hello

The service is still available because its PodDisruptionBudget is respected.

WatchedEvent state:SyncConnected type:None path:null
world
cZxid = 0x200000002
ctime = Wed Dec 07 00:08:59 UTC 2016
mZxid = 0x200000002
mtime = Wed Dec 07 00:08:59 UTC 2016
pZxid = 0x200000002
cversion = 0
dataVersion = 0
aclVersion = 0
ephemeralOwner = 0x0
dataLength = 5
numChildren = 0

Use kubectl uncordon to uncordon the first node.

kubectl uncordon kubernetes-node-pb41

The output is similar to this:

node "kubernetes-node-pb41" uncordoned

zk-1 is rescheduled on this node. Wait until zk-1 is Running and Ready.

kubectl get pods -w -l app=zk

The output is similar to this:

NAME      READY     STATUS    RESTARTS   AGE
zk-0      1/1       Running   2          1h
zk-1      1/1       Running   0          1h
zk-2      1/1       Running   0          1h
NAME      READY     STATUS        RESTARTS   AGE
zk-0      1/1       Terminating   2          2h
zk-0      0/1       Terminating   2         2h
zk-0      0/1       Terminating   2         2h
zk-0      0/1       Terminating   2         2h
zk-0      0/1       Pending   0         0s
zk-0      0/1       Pending   0         0s
zk-0      0/1       ContainerCreating   0         0s
zk-0      0/1       Running   0         51s
zk-0      1/1       Running   0         1m
zk-1      1/1       Terminating   0         2h
zk-1      0/1       Terminating   0         2h
zk-1      0/1       Terminating   0         2h
zk-1      0/1       Terminating   0         2h
zk-1      0/1       Pending   0         0s
zk-1      0/1       Pending   0         0s
zk-1      0/1       Pending   0         12m
zk-1      0/1       ContainerCreating   0         12m
zk-1      0/1       Running   0         13m
zk-1      1/1       Running   0         13m

Attempt to drain the node on which zk-2 is scheduled.

kubectl drain $(kubectl get pod zk-2 --template {{.spec.nodeName}}) --ignore-daemonsets --force --delete-emptydir-data

The output is similar to this:

node "kubernetes-node-i4c4" already cordoned
WARNING: Deleting pods not managed by ReplicationController, ReplicaSet, Job, or DaemonSet: fluentd-cloud-logging-kubernetes-node-i4c4, kube-proxy-kubernetes-node-i4c4; Ignoring DaemonSet-managed pods: node-problem-detector-v0.1-dyrog
pod "heapster-v1.2.0-2604621511-wht1r" deleted
pod "zk-2" deleted
node "kubernetes-node-i4c4" drained

This time kubectl drain succeeds.

Uncordon the second node to allow zk-2 to be rescheduled.

kubectl uncordon kubernetes-node-ixsl

The output is similar to this:

node "kubernetes-node-ixsl" uncordoned

You can use kubectl drain in conjunction with PodDisruptionBudgets to ensure that your services remain available during maintenance. If drain is used to cordon nodes and evict pods prior to taking the node offline for maintenance, services that express a disruption budget will have that budget respected. You should always allocate additional capacity for critical services so that their Pods can be immediately rescheduled.

Cleaning up

  • Use kubectl uncordon to uncordon all the nodes in your cluster.
  • You must delete the persistent storage media for the PersistentVolumes used in this tutorial. Follow the necessary steps, based on your environment, storage configuration, and provisioning method, to ensure that all storage is reclaimed.