https://github.com/jinghzhu/kubernetespodoperator
Example of Controller for Pod resources in Kubernetes
https://github.com/jinghzhu/kubernetespodoperator
go kubernetes operator pod
Last synced: 7 months ago
JSON representation
Example of Controller for Pod resources in Kubernetes
- Host: GitHub
- URL: https://github.com/jinghzhu/kubernetespodoperator
- Owner: jinghzhu
- License: mit
- Created: 2018-08-13T09:20:41.000Z (about 7 years ago)
- Default Branch: master
- Last Pushed: 2023-02-06T23:52:33.000Z (over 2 years ago)
- Last Synced: 2025-01-27T13:09:54.203Z (8 months ago)
- Topics: go, kubernetes, operator, pod
- Language: Go
- Size: 20.8 MB
- Stars: 2
- Watchers: 2
- Forks: 1
- Open Issues: 1
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Kubernetes CRD Controller
This repository is to provide a sample about how to create a controller to watch all events (add/update/delete) of Pods so that we can implement the Operator pattern.
I develop it to work with followings together to demo a sample [ReplicaSet](https://kubernetes.io/docs/concepts/workloads/controllers/replicaset/) scenario:
1. CRD [Jinghzhu v1](https://github.com/jinghzhu/KubernetesCRD)
2. [CRD Controller](https://github.com/jinghzhu/KubernetescrdOperator/)
Assume:
1. Go version > 1.9.0
2. A Kubernetes cluster is available.
3. Kubernetes version >= 1.18.0
# How It Works
There are many articles introducing the mechanism of controller in Kubernetes community. So, I wouldn't pay too much attention on it. In short, it leverage the etcd watch mechanism to catch all events.
In the method `New()` at `pkg/operator/types.go`, we define the major components of Operator:
1. queue - I implements an in-memory queue to process all events in concurrent way. By default, there is only one main goroutine to help the Operator get events from etcd and call add/update/delete handlers. Now, after receiving events, it puts them in the queue which can ensure the concurrent safe.
2. lister
3. informer
4. event handler callback
```go
op := &Operator{
kubeClient: client,
podsQueue: workqueue.NewNamedRateLimitingQueue(workqueue.NewItemExponentialFailureRateLimiter(100*time.Millisecond, 5*time.Second), "pods"),
context: ctx,
}
// fieldSelector := labels.Set{"keys": string(nodeName)}.AsSelector()
ListOption := metav1.ListOptions{}
op.indexer, op.watcher = cache.NewIndexerInformer(
// cache.NewListWatchFromClient(client.CoreV1().RESTClient(), "pods", namespace, fieldSelector),
&cache.ListWatch{
ListFunc: func(options metav1.ListOptions) (runtime.Object, error) {
options = ListOption
return op.kubeClient.CoreV1().Pods(namespace).List(ctx, options)
},
WatchFunc: func(options metav1.ListOptions) (watch.Interface, error) {
options = ListOption
return op.kubeClient.CoreV1().Pods(namespace).Watch(ctx, options)
},
},
&corev1.Pod{},
0,
cache.ResourceEventHandlerFuncs{
AddFunc: op.onAdd,
UpdateFunc: op.onUpdate,
DeleteFunc: op.onDelete,
},
cache.Indexers{},
)
```
For details, please see `pkg/operator/operator.go` and `pkg/operator/watcher.go`.
# How It Looks Like
```bash
$ go run cmd/operator/main.go
Init Pod Operator...
Ready to new an Operator object for kubeconfig /home/jinghzhu/.kube/config and namespace worker
Ready to start Pod Operator...
Pod Operator is running...
Find a Pod add event.
Name: jinghzhu-worker-hl7fc
Resource Version: 2347337
```