如何为类型配置自动缩放：对象

Posted 2023-03-04

【中文标题】如何为类型配置自动缩放：对象

【英文标题】：How to configure autoscaling for type: Object

【发布时间】：2019-03-14 14:27:18

【问题描述】：

我配置了一个自定义指标。

当我跑步时：

kubectl get --raw "/apis/custom.metrics.k8s.io/v1beta1/namespaces/my-namespace/deployments.extensions/sample-app/queue_length" | jq

我明白了：

...
  "items": [
    
      "describedObject": 
        "kind": "Deployment",
        "namespace": "my-namespace",
        "name": "sample-app",
        "apiVersion": "extensions/v1beta1"
      ,
      "metricName": "queue_length",
      "timestamp": "2018-10-09T18:14:15Z",
      "value": "30"
    
  ]
...

所以我想用这个指标来扩展我的部署：

kind: HorizontalPodAutoscaler
apiVersion: autoscaling/v2beta1
metadata:
  name: sample-app
  namespace: my-namespace
spec:
  scaleTargetRef:
    apiVersion: extensions/v1beta1
    kind: Deployment
    name: sample-app
  minReplicas: 1
  maxReplicas: 10
  metrics:
  - type: Object
    object:
      metricName: queue_length
      target:
        apiVersion: extensions/v1beta1
        kind: Deployment
        name: sample-app
      targetValue: 20

但这似乎不起作用，也许我必须指定- type: Object有点不同。

我尝试将kind 更改为部署、部署.扩展、部署.扩展等，但我不断收到错误FailedGetScale

更新：我发现 HPA 对象和目标对象的命名空间应该相同，所以我更改了它。现在我得到了

the HPA was unable to compute the replica count: unable to get metric queue_length: Deployment on default sample-app/unable to fetch metrics from custom metrics API: the server could not find the metric queue_length for deployments.extensions sample-app

Upd2.: 弄清楚了：hpa 在命名空间 default 中，而 metric 暴露在 my-namespace 中。所以，应该小心命名空间！

【参考方案1】：

试试这个：

---
kind: HorizontalPodAutoscaler
apiVersion: autoscaling/v2beta1
metadata:
  name: sample-app
  namespace: my-namespace
spec:
  scaleTargetRef:
    apiVersion: extensions/v1beta1
    kind: Deployment
    name: sample-app
  minReplicas: 1
  maxReplicas: 10
  metrics:
  - type: Object
    object:
      target:
        kind: Service
        name: sample-app
      metricName: queue_length
      targetValue: 20

【讨论】：

试过了，但也不管用：ScalingActive False FailedGetObjectMetric the HPA was unable to compute the replica count: unable to get metric celery_production_yodlee: Service on default sample-app/unable to fetch metrics from custom metrics API: the server could not find the metric celery_production_yodlee for services.extensions。因为我有 API 路径 namespaces/my-namespace/deployments.extensions/sample-app/queue_length 应该有 deployment 我认为

哦，糟糕。什么是回归？ kubectl get --raw "/apis/custom.metrics.k8s.io/v1beta1/namespaces/my-namespace/services/*/queue_length" | jq .