DaemonSet Upgrade Model
Background
In edge scenarios, the native DaemonSet upgrade model does not perfectly satisfy existing requirements. In the case of cloud-edge network disconnection, DaemonSet upgrade process may be blocked. In addition, the native upgrade model does not provide any upgrade operation API, and users cannot control the application upgrade on their own.
In order to address the above problems, we extend the native DaemonSet upgrade model by adding a custom controller daemonPodUpdater-controller
, providing Auto and OTA two upgrade model.
- Auto: Solve the DaemonSet upgrade process blocking problem which caused by node
Not-Ready
when the cloud-edge is disconnected. During auto upgrade,not-ready
nodes will be ignored. And whenNot-Ready
nodes turn toReady
, upgrade process will be completed automatically. - OTA: Add pod status condition
PodNeedUpgrade
which indicates the upgrade availability information. YurtHub OTA component can use this condition to determine if a new version of DaemonSet application exists.
Configuration
# example configuration for auto or ota upgrade
apiVersion: apps/v1
kind: DaemonSet
metadata:
# ···
annotations:
# This annotation is the first prerequisite for using auto or ota upgrade
# and the only valid values are "auto" or "ota".
apps.openyurt.io/update-strategy: ota
# This annotation is used for rolling update and only works in auto mode.
# The supported value is the same with native DaemonSet maxUnavailable, default to 10%.
apps.openyurt.io/max-unavailable: 30%
# ···
spec:
# ···
# Set updateStrategy to "OnDelete" is another prerequisite for using auto or ota upgrade.
updateStrategy:
type: OnDelete
# ···
In short, if you wish to use Auto or OTA upgrade, you need to set annotation apps.openyurt.io/update-strategy
to “auto” or “ota” and set .spec.updateStrategy.type
to “OnDelete”.
Usage:
1)Install Yurt-Controller-Manager Components
daemonPodUpdater-controller
is integrated in Yurt-Controller-Manager
, and it needs to be installed before using Auto or OTA Upgrade Model, you can refer to Deploying OpenYurt Components
2)Auto Upgrade Model
- Create daemonset instance
cat <<EOF | kubectl apply -f -
apiVersion: apps/v1
kind: DaemonSet
metadata:
name: nginx-daemonset
annotations:
apps.openyurt.io/update-strategy: auto
spec:
selector:
matchLabels:
app: nginx
updateStrategy:
type: OnDelete
template:
metadata:
labels:
app: nginx
spec:
containers:
- name: nginx
image: nginx:1.19.4
EOF
- Get nginx-daemonset pods
$ kubectl get pods | grep nginx-daemonset
nginx-daemonset-bv5jg 1/1 Running 0 21m 10.244.2.2 openyurt-e2e-test-worker3 <none> <none>
nginx-daemonset-fhsr6 1/1 Running 0 21m 10.244.1.2 openyurt-e2e-test-worker <none> <none>
nginx-daemonset-lmmtd 1/1 Running 0 21m 10.244.3.2 openyurt-e2e-test-worker2 <none> <none>
- Simulate cloud-edge network disconnection: assume that nodes
openyurt-e2e-test-worker2
andopenyurt-e2e-test-worker3
are disconnected from the cloud node. This example usesKind
to create the cluster, so the network disconnection can be simulated by removing the containers from the virtual bridge.
$ docker network disconnect kind openyurt-e2e-test-worker2
$ docker network disconnect kind openyurt-e2e-test-worker3
$ kubectl get nodes -o wide
AME STATUS ROLES AGE VERSION INTERNAL-IP EXTERNAL-IP OS-IMAGE KERNEL-VERSION CONTAINER-RUNTIME
openyurt-e2e-test-control-plane Ready control-plane,master 36m v1.22.7 172.18.0.4 <none> Ubuntu 21.10 5.10.76-linuxkit containerd://1.5.10
openyurt-e2e-test-worker Ready <none> 35m v1.22.7 172.18.0.2 <none> Ubuntu 21.10 5.10.76-linuxkit containerd://1.5.10
openyurt-e2e-test-worker2 NotReady <none> 35m v1.22.7 172.18.0.3 <none> Ubuntu 21.10 5.10.76-linuxkit containerd://1.5.10
openyurt-e2e-test-worker3 NotReady <none> 35m v1.22.7 172.18.0.5 <none> Ubuntu 21.10 5.10.76-linuxkit containerd://1.5.10
- Update daemonset: change the container image from nginx:1.19.4 to nginx:1.19.5
***
containers:
- name: nginx
image: nginx:1.19.5
***
- Get pods: the old pod
default/nginx-daemonset-fhsr6
onopenyurt-e2e-test-worker
node has been deleted and the new poddefault/nginx-daemonset-slp5t
has been created; the pods on the two disconnected nodes will not be upgraded temporarily
nginx-daemonset-bv5jg 1/1 Running 0 33m 10.244.2.2 openyurt-e2e-test-worker3 <none> <none>
nginx-daemonset-lmmtd 1/1 Running 0 33m 10.244.3.2 openyurt-e2e-test-worker2 <none> <none>
nginx-daemonset-slp5t 1/1 Running 0 5m54s 10.244.1.3 openyurt-e2e-test-worker <none> <none>
- Restore network connectivity of nodes
$ docker network connect kind openyurt-e2e-test-worker2
$ docker network connect kind openyurt-e2e-test-worker3
$ kubectl get nodes -o wide
NAME STATUS ROLES AGE VERSION INTERNAL-IP EXTERNAL-IP OS-IMAGE KERNEL-VERSION CONTAINER-RUNTIME
openyurt-e2e-test-control-plane Ready control-plane,master 49m v1.22.7 172.18.0.4 <none> Ubuntu 21.10 5.10.76-linuxkit containerd://1.5.10
openyurt-e2e-test-worker Ready <none> 48m v1.22.7 172.18.0.2 <none> Ubuntu 21.10 5.10.76-linuxkit containerd://1.5.10
openyurt-e2e-test-worker2 Ready <none> 48m v1.22.7 172.18.0.3 <none> Ubuntu 21.10 5.10.76-linuxkit containerd://1.5.10
openyurt-e2e-test-worker3 Ready <none> 48m v1.22.7 172.18.0.5 <none> Ubuntu 21.10 5.10.76-linuxkit containerd://1.5.10
- Get pods: daemonset pods on all nodes have been upgraded
nginx-daemonset-kbkf6 1/1 Running 0 88s 10.244.3.3 openyurt-e2e-test-worker2 <none> <none>
nginx-daemonset-scgtv 1/1 Running 0 51s 10.244.2.3 openyurt-e2e-test-worker3 <none> <none>
nginx-daemonset-slp5t 1/1 Running 0 11m 10.244.1.3 openyurt-e2e-test-worker <none> <none>
- Check pods image version: all pods have been upgraded to nginx:1.19.5
***
Containers:
nginx:
Container ID: containerd://f7d4b3f1257a0d1d8da862671c11cb094f9fba1ba0041b7a5f783d9c9e4d8449
Image: nginx:1.19.5
Image ID: docker.io/library/nginx@sha256:31de7d2fd0e751685e57339d2b4a4aa175aea922e592d36a7078d72db0a45639
Port: <none>
Host Port: <none>
State: Running
Started: Fri, 14 Oct 2022 14:21:25 +0800
Ready: True
Restart Count: 0
Environment: <none>
Mounts:
/var/run/secrets/kubernetes.io/serviceaccount from kube-api-access-wrhj8 (ro)
***
3)OTA Upgrade Model
OTA Upgrade API
YurtHub provides two REST APIs for OTA upgrades.
GET /pods
This API allows you to get information about the pods on the node.
POST /openyurt.io/v1/namespaces/{ns}/pods/{podname}/upgrade
This API allows you to specify and upgrade a DaemonSet Pod. The path parameters
ns
andpodname
represent the namespace and name of the pod, respectively.
OTA Upgrade Example
- Create daemonset instance
cat <<EOF | kubectl apply -f -
apiVersion: apps/v1
kind: DaemonSet
metadata:
name: nginx-daemonset
annotations:
apps.openyurt.io/update-strategy: ota
spec:
selector:
matchLabels:
app: nginx
updateStrategy:
type: OnDelete
template:
metadata:
labels:
app: nginx
spec:
containers:
- name: nginx
image: nginx:1.19.4
EOF
# get nginx-daemonset pods
$ kubectl get pods -o wide | grep nginx-daemonset
nginx-daemonset-bwzss 1/1 Running 0 92s 10.244.3.4 openyurt-e2e-test-worker2 <none> <none>
nginx-daemonset-ppf9p 1/1 Running 0 92s 10.244.1.4 openyurt-e2e-test-worker <none> <none>
nginx-daemonset-rgp9h 1/1 Running 0 92s 10.244.2.4 openyurt-e2e-test-worker3 <none> <none>
- Check pod status condition
PodNeedUpgrade
: take podnginx-daemonset-bwzss
on nodeopenyurt-e2e-test-worker2
as an example
$ kubectl describe pods nginx-daemonset-bwzss
***
Conditions:
Type Status
PodNeedUpgrade False
***
- Update daemonset: change the container image from nginx:1.19.4 to nginx:1.19.5
***
containers:
- name: nginx
image: nginx:1.19.5
***
- Check pod status condition
PodNeedUpgrade
again
$ kubectl describe pods nginx-daemonset-bwzss
***
Conditions:
Type Status
PodNeedUpgrade True
***
- Execute OTA upgrade
# enter edge node container of Kind cluster
$ docker exec -it openyurt-e2e-test-worker2 /bin/bash
# call Upgrade API, this upgrade API is only available on the edge nodes
$ curl -X POST 127.0.0.1:10267/openyurt.io/v1/namespaces/default/pods/nginx-daemonset-bwzss/upgrade
Start updating pod default/nginx-daemonset-bwzss
- Check upgrade result: pod
nginx-daemonset-bwzss
on nodeopenyurt-e2e-test-worker2
has been deleted and new podnginx-daemonset-vrvhn
has been created
# check result
$ kubectl get pods -o wide | grep nginx-daemonset
nginx-daemonset-ppf9p 1/1 Running 0 15m 10.244.1.4 openyurt-e2e-test-worker <none> <none>
nginx-daemonset-rgp9h 1/1 Running 0 15m 10.244.2.4 openyurt-e2e-test-worker3 <none> <none>
nginx-daemonset-vrvhn 1/1 Running 0 63s 10.244.3.5 openyurt-e2e-test-worker2 <none> <none>
# check pod container image
$ kubectl describe pods nginx-daemonset-vrvhn
***
Containers:
nginx:
Container ID: containerd://18df6aa88076639353ea0b3d87f340cd4c86ab27a7f154bce06345e9764c997a
Image: nginx:1.19.5
Image ID: docker.io/library/nginx@sha256:31de7d2fd0e751685e57339d2b4a4aa175aea922e592d36a7078d72db0a45639
Port: <none>
Host Port: <none>
State: Running
Started: Fri, 14 Oct 2022 16:25:20 +0800
Ready: True
Restart Count: 0
Environment: <none>
Mounts:
/var/run/secrets/kubernetes.io/serviceaccount from kube-api-access-p6kjh (ro)
***