模拟Pod不可用
介绍
kubernetes Pod 资源自身场景, 所选的 Pod 在指定的时间段内将不可用
命令
blade create k8s pod-pod fail -h
参数
参数名 | 说明 | 类型 | 值 |
---|---|---|---|
evict-count | 限制实验生效的数量 | int | |
evict-percent | 限制实验生效数量的百分比,不包含 % | int | |
namespace | Pod 所属的命名空间,只能填写一个值,必填项 | string | 例:default |
labels | Pod 资源标签,多个标签之间是或的关系 | string | |
names | Pod 资源名 | string | |
kubeconfig | kubeconfig 文件全路径(仅限使用 blade 命令调用时使用) | string | 例: "/root/.kube/config" |
waiting-time | 实验结果等待时间,默认为 20s,参数值要包含单位,例如 10s,1m | string |
案例
使指定 default 命名空间下标签是 app=guestbook 的 pod 服务不可用
yaml配置方式如下
apiVersion: chaosblade.io/v1alpha1
kind: ChaosBlade
metadata:
name: fail-pod-by-labels
spec:
experiments:
- scope: pod
target: pod
action: fail
desc: "inject fail image to select pod"
matchers:
- name: labels
value:
- "app=guestbook"
- name: namespace
value:
- "default"
保存文件为 fail_pod_by_labels.yaml,使用 kubectl apply -f fail_pod_by_labels.yaml
命令执行,可以看到执行前后,指定的 Pod 无法拉取镜像,删除实验之后,Pod 又恢复正常
frontend-5d944bddbb-8w9ff 1/1 Running 0 45h
frontend-5d944bddbb-8w9ff 1/1 Running 0 45h
frontend-5d944bddbb-8w9ff 0/1 ErrImagePull 0 45h
frontend-5d944bddbb-8w9ff 0/1 ImagePullBackOff 0 45h
frontend-5d944bddbb-8w9ff 0/1 ErrImagePull 0 45h
frontend-5d944bddbb-8w9ff 0/1 ImagePullBackOff 0 45h
frontend-5d944bddbb-8w9ff 0/1 CrashLoopBackOff 0 45h
frontend-5d944bddbb-8w9ff 0/1 ErrImagePull 0 45h
frontend-5d944bddbb-8w9ff 0/1 Terminating 0 45h
frontend-5d944bddbb-8sll8 0/1 Pending 0 0s
frontend-5d944bddbb-8sll8 0/1 Pending 0 0s
frontend-5d944bddbb-8w9ff 0/1 Terminating 0 45h
frontend-5d944bddbb-8sll8 0/1 ContainerCreating 0 0s
frontend-5d944bddbb-8sll8 1/1 Running 0 4s
frontend-5d944bddbb-8w9ff 0/1 Terminating 0 45h
frontend-5d944bddbb-8w9ff 0/1 Terminating 0 45h
通过 kubectl get blade fail-pod-by-labels -o json
可以查看详细的执行结果(下发只截取部分内容)
{
"apiVersion": "chaosblade.io/v1alpha1",
"kind": "ChaosBlade",
"metadata": {
"creationTimestamp": "2020-07-05T01:18:01Z",
"finalizers": [
"finalizer.chaosblade.io"
],
"generation": 1,
"name": "fail-pod-by-labels",
"resourceVersion": "149831443",
"selfLink": "/apis/chaosblade.io/v1alpha1/chaosblades/fail-pod-by-labels",
"uid": "5e7c42f8-be5d-11ea-b2c3-00163f003dea"
},
"spec": {
"experiments": [
{
"action": "fail",
"desc": "inject fail image to select pod",
"matchers": [
{
"name": "labels",
"value": [
"app=guestbook"
]
},
{
"name": "namespace",
"value": [
"default"
]
}
],
"scope": "pod",
"target": "pod"
}
]
},
"status": {
"expStatuses": [
{
"action": "fail",
"resStatuses": [
{
"kind": "pod",
"name": "frontend-5d944bddbb-8w9ff",
"nodeName": "qa.op.k8s.008120.hz",
"state": "Success",
"success": true,
"uid": "5187a405-be5d-11ea-a750-00163e0dc939"
}
],
"scope": "pod",
"state": "Success",
"success": true,
"target": "pod"
}
],
"phase": "Running"
}
}
执行以下命令停止实验:
kubectl delete -f fail_pod_by_labels.yaml
或者直接删除 blade 资源:
kubectl delete blade fail-pod-by-labels
Pod 不可用的停止实验操作,chaosblade 会重新拉起被删除的 Pod
常见问题
其他问题参考 blade create k8s 常见问题