Skip to content

Latest commit

 

History

History
159 lines (122 loc) · 6.33 KB

advanced-daemonset.md

File metadata and controls

159 lines (122 loc) · 6.33 KB
Raw

Use Advanced DaemonSet to deploy daemons

This tutorial goes through an example of deploying and updating daemons using Advanced DaemonSet.

Prepare nodes

In this example, the testing cluster has five available nodes. In order to verify node selector update, we add a node-type=canary label to one of them.

$ kubectl label node cn-hangzhou.192.168.0.10 node-type=canary --overwrite
node/cn-hangzhou.192.168.0.10 labeled

$ kubectl get nodes -L node-type
NAME                       STATUS   ROLES    AGE   VERSION            NODE-TYPE
cn-hangzhou.192.168.0.10   Ready    <none>   89d   v1.16.9-aliyun.1   canary
cn-hangzhou.192.168.0.11   Ready    <none>   89d   v1.16.9-aliyun.1
cn-hangzhou.192.168.0.12   Ready    <none>   89d   v1.16.9-aliyun.1
cn-hangzhou.192.168.0.23   Ready    <none>   58d   v1.16.9-aliyun.1
cn-hangzhou.192.168.0.24   Ready    <none>   58d   v1.16.9-aliyun.1

Install guestbook DaemonSet

$ kubectl apply -f https://raw.githubusercontent.com/kruiseio/kruise/master/docs/tutorial/v1/daemonset-guestbook.yaml
daemonset.apps.kruise.io/guestbook-ds created

Check to make sure all Pods are running and ready.

$ kubectl get pod -L controller-revision-hash -l app=guestbook-daemon -o wide
NAME                 READY   STATUS    RESTARTS   AGE   IP             NODE                       CONTROLLER-REVISION-HASH
guestbook-ds-bwtl4   1/1     Running   0          8s    172.27.0.186   cn-hangzhou.192.168.0.11   864fbf5949
guestbook-ds-fbhnn   1/1     Running   0          8s    172.27.1.31    cn-hangzhou.192.168.0.24   864fbf5949
guestbook-ds-fvq49   1/1     Running   0          8s    172.27.0.62    cn-hangzhou.192.168.0.12   864fbf5949
guestbook-ds-mbd56   1/1     Running   0          8s    172.27.0.110   cn-hangzhou.192.168.0.10   864fbf5949
guestbook-ds-t8dd8   1/1     Running   0          8s    172.27.0.217   cn-hangzhou.192.168.0.23   864fbf5949

Update guestbook with Standard methodology

Using selector

Edit the DaemonSet, update image to openkruise/guestbook:v2 and set node selector node-type: canary in the rollingUpdate spec:

$ kubectl edit daemonset.apps.kruise.io guestbook-ds
...

  updateStrategy:
    rollingUpdate:
      selector:
        matchLabels:
          node-type: canary
      maxUnavailable: 1
      partition: 0
      rollingUpdateType: Standard
    type: RollingUpdate

Then we will find the Pod on cn-hangzhou.192.168.0.10, which has been added the canary label, has been recreated using the new version.

$ kubectl get pod -L controller-revision-hash -l app=guestbook-daemon -o wide
NAME                 READY   STATUS    RESTARTS   AGE   IP             NODE                       CONTROLLER-REVISION-HASH
guestbook-ds-bwtl4   1/1     Running   0          73s   172.27.0.186   cn-hangzhou.192.168.0.11   864fbf5949
guestbook-ds-fbhnn   1/1     Running   0          73s   172.27.1.31    cn-hangzhou.192.168.0.24   864fbf5949
guestbook-ds-fvq49   1/1     Running   0          73s   172.27.0.62    cn-hangzhou.192.168.0.12   864fbf5949
guestbook-ds-t8dd8   1/1     Running   0          73s   172.27.0.217   cn-hangzhou.192.168.0.23   864fbf5949
guestbook-ds-xnld8   1/1     Running   0          3s    172.27.0.111   cn-hangzhou.192.168.0.10   cdf6d4478

Using partition

Edit the DaemonSet, remove the node selector and set partition to 2.

$ kubectl edit daemonset.apps.kruise.io guestbook-ds
...

  updateStrategy:
    rollingUpdate:
      maxUnavailable: 1
      partition: 2
      rollingUpdateType: Standard
    type: RollingUpdate

Then the controller will update two more Pods one by one, to reach the expected 5 - 2 = 3 Pods to use the new version in addition to the one that has been updated in previous step.

$ kubectl get pod -L controller-revision-hash -l app=guestbook-daemon -o wide
NAME                 READY   STATUS    RESTARTS   AGE    IP             NODE                       CONTROLLER-REVISION-HASH
guestbook-ds-bwtl4   1/1     Running   0          2m5s   172.27.0.186   cn-hangzhou.192.168.0.11   864fbf5949
guestbook-ds-fvq49   1/1     Running   0          2m5s   172.27.0.62    cn-hangzhou.192.168.0.12   864fbf5949
guestbook-ds-pxhn4   1/1     Running   0          11s    172.27.1.32    cn-hangzhou.192.168.0.24   cdf6d4478
guestbook-ds-txqjl   1/1     Running   0          2s     172.27.0.218   cn-hangzhou.192.168.0.23   cdf6d4478
guestbook-ds-xnld8   1/1     Running   0          55s    172.27.0.111   cn-hangzhou.192.168.0.10   cdf6d4478

Updating the rest of Pods

Edit the DaemonSet and set partition to 0.

$ kubectl edit daemonset.apps.kruise.io guestbook-ds
...

  updateStrategy:
    rollingUpdate:
      maxUnavailable: 1
      partition: 0
      rollingUpdateType: Standard
    type: RollingUpdate

All Pods will be updated.

$ kubectl get pod -L controller-revision-hash -l app=guestbook-daemon -o wide
NAME                 READY   STATUS    RESTARTS   AGE     IP             NODE                       CONTROLLER-REVISION-HASH
guestbook-ds-j76db   1/1     Running   0          3s      172.27.0.7     cn-hangzhou.192.168.0.12   cdf6d4478
guestbook-ds-lcbnj   1/1     Running   0          14s     172.27.0.187   cn-hangzhou.192.168.0.11   cdf6d4478
guestbook-ds-pxhn4   1/1     Running   0          7m34s   172.27.1.32    cn-hangzhou.192.168.0.24   cdf6d4478
guestbook-ds-txqjl   1/1     Running   0          7m25s   172.27.0.218   cn-hangzhou.192.168.0.23   cdf6d4478
guestbook-ds-xnld8   1/1     Running   0          8m18s   172.27.0.111   cn-hangzhou.192.168.0.10   cdf6d4478

Update guestbook with Surging methodology

The rollingUpdateType could be Standard or Surging.

maxUnavailable is valid for Standard type and maxSurge is valid for Surging type.

  updateStrategy:
    rollingUpdate:
      maxSurge: 1
      partition: 0
      rollingUpdateType: Surging
    type: RollingUpdate

When rollingUpdateType is Surging:

  1. For each node, the controller will first create a new Pod, and then delete the old one when the new one has been ready.
  2. maxSurge controls how many nodes could be updated at one time.

Go and try Surging by yourself to understand how it works.

Uninstall guestbook DaemonSet

$ kubectl delete -f https://raw.githubusercontent.com/kruiseio/kruise/master/docs/tutorial/v1/daemonset-guestbook.yaml
daemonset.apps.kruise.io "guestbook-ds" deleted