Kubernetes不会在GKE上触发自动缩放

出于某种原因，Kubernetes 1.6.2不会在Google容器引擎上触发自动缩放

我有一个

someservice

定义，包含以下资源和滚动更新：

apiVersion: extensions/v1beta1
kind: Deployment
metadata:
  name: someservice
  labels:
    layer: backend
spec:
  minReadySeconds: 160
  replicas: 1
  strategy:
    rollingUpdate:
      maxSurge: 100%
      maxUnavailable: 0
    type: RollingUpdate  
  template:
    metadata:
      labels:
        name: someservice
        layer: backend        
    spec:
      containers:
      - name: someservice
        image: eu.gcr.io/XXXXXX/someservice:v1
        imagePullPolicy: Always
        resources:
          limits:
            cpu: 2
            memory: 20Gi        
          requests:
            cpu: 400m
            memory: 18Gi
     <.....>

我的节点池设置为自动缩放，最小值为2，最大值为5。节点池中的机器（

n1-highmem-8

）足够大（52GB）来容纳此服务。但不知何故，什么也没有发生：

$ kubectl get nodes
NAME                                 STATUS    AGE       VERSION
gke-dev-default-pool-efca0068-4qq1   Ready     2d        v1.6.2
gke-dev-default-pool-efca0068-597s   Ready     2d        v1.6.2
gke-dev-default-pool-efca0068-6srl   Ready     2d        v1.6.2
gke-dev-default-pool-efca0068-hb1z   Ready     2d        v1.6.2

$ kubectl  describe nodes | grep -A 4 'Allocated resources'
Allocated resources:
  (Total limits may be over 100 percent, i.e., overcommitted.)
  CPU Requests  CPU Limits  Memory Requests     Memory Limits
  ------------  ----------  ---------------     -------------
  7060m (88%)   15510m (193%)   39238591744 (71%)   48582818048 (88%)
--
Allocated resources:
  (Total limits may be over 100 percent, i.e., overcommitted.)
  CPU Requests  CPU Limits  Memory Requests Memory Limits
  ------------  ----------  --------------- -------------
  6330m (79%)   22200m (277%)   48930Mi (93%)   66344Mi (126%)
--
Allocated resources:
  (Total limits may be over 100 percent, i.e., overcommitted.)
  CPU Requests  CPU Limits  Memory Requests Memory Limits
  ------------  ----------  --------------- -------------
  7360m (92%)   13200m (165%)   49046Mi (93%)   44518Mi (85%)
--
Allocated resources:
  (Total limits may be over 100 percent, i.e., overcommitted.)
  CPU Requests  CPU Limits  Memory Requests     Memory Limits
  ------------  ----------  ---------------     -------------
  7988m (99%)   11538m (144%)   32967256Ki (61%)    21690968Ki (40%)

$ gcloud container node-pools describe  default-pool --cluster=dev
autoscaling:
  enabled: true
  maxNodeCount: 5
  minNodeCount: 2
config:
  diskSizeGb: 100
  imageType: COS
  machineType: n1-highmem-8
  oauthScopes:
  - https://www.googleapis.com/auth/compute
  - https://www.googleapis.com/auth/datastore
  - https://www.googleapis.com/auth/devstorage.read_only
  - https://www.googleapis.com/auth/devstorage.read_write
  - https://www.googleapis.com/auth/service.management.readonly
  - https://www.googleapis.com/auth/servicecontrol
  - https://www.googleapis.com/auth/sqlservice
  - https://www.googleapis.com/auth/logging.write
  - https://www.googleapis.com/auth/monitoring
  serviceAccount: default
initialNodeCount: 2
instanceGroupUrls:
- https://www.googleapis.com/compute/v1/projects/XXXXXX/zones/europe-west1-b/instanceGroupManagers/gke-dev-default-pool-efca0068-grp
management:
  autoRepair: true
name: default-pool
selfLink: https://container.googleapis.com/v1/projects/XXXXXX/zones/europe-west1-b/clusters/dev/nodePools/default-pool
status: RUNNING
version: 1.6.2

$ kubectl -n dev get pods -l name=someservice
NAME                      READY     STATUS    RESTARTS   AGE
someservice-2595684989-h8c5d   0/1       Pending   0          42m
someservice-804061866-f2trc    1/1       Running   0          1h

$ kubectl -n dev describe pod someservice-2595684989-h8c5d

Events:
  FirstSeen LastSeen    Count   From            SubObjectPath   Type        Reason          Message
  --------- --------    -----   ----            -------------   --------    ------          -------
  43m       43m     4   default-scheduler           Warning     FailedScheduling    No nodes are available that match all of the following predicates:: Insufficient cpu (4), Insufficient memory (3).
  43m       42m     6   default-scheduler           Warning     FailedScheduling    No nodes are available that match all of the following predicates:: Insufficient cpu (3), Insufficient memory (3).
  41m       41m     2   default-scheduler           Warning     FailedScheduling    No nodes are available that match all of the following predicates:: Insufficient cpu (2), Insufficient memory (3).
  40m       36s     136 default-scheduler           Warning     FailedScheduling    No nodes are available that match all of the following predicates:: Insufficient cpu (1), Insufficient memory (3).
  43m       2s      243 cluster-autoscaler          Normal      NotTriggerScaleUp   pod didn't trigger scale-up (it wouldn't fit if a new node is added)


$ kubectl version
Client Version: version.Info{Major:"1", Minor:"6", GitVersion:"v1.6.2", GitCommit:"477efc3cbe6a7effca06bd1452fa356e2201e1ee", GitTreeState:"clean", BuildDate:"2017-04-19T20:33:11Z", GoVersion:"go1.7.5", Compiler:"gc", Platform:"darwin/amd64"}
Server Version: version.Info{Major:"1", Minor:"6", GitVersion:"v1.6.2", GitCommit:"477efc3cbe6a7effca06bd1452fa356e2201e1ee", GitTreeState:"clean", BuildDate:"2017-04-19T20:22:08Z", GoVersion:"go1.7.5", Compiler:"gc", Platform:"linux/amd64"}

---

看来这是Kubernetes 1.6.2的一个bug。根据GKE支持工程师的说法：

从消息“没有与所有节点匹配的节点可用” 在谓词之后，“这似乎是一个已知的问题 工程师们设法找到了根本原因。这是我的一个问题 GKE 1.6（上一版本）中当前使用的群集自动缩放器版本0.5.1 至1.6.2）。此问题已在群集autoscaler中修复 0.5.2，包含在1.6分支的水头中

---

确保实例组autoscaler已禁用或具有适当的最小/最大实例数设置

根据：

基于CPU（或任何基于度量的）群集/节点组自动缩放器，如 GCE实例组Autoscaler与[Kubernetes]不兼容 群集奥斯卡勒]。它们也不是特别适合使用 和库伯内特斯在一起

…所以它可能应该被禁用

尝试：

然后检查

autoscaler

属性。如果禁用实例组autoscaler，则将不存在该选项

要禁用它，请执行以下操作：

gcloud compute instance-groups managed stop-autoscaling gke-dev-default-pool-efca0068-grp \
    --zone europe-west1-b

gcloud compute instance-groups managed describe gke-dev-default-pool-efca0068-grp \
    --zone europe-west1-b

gcloud compute instance-groups managed stop-autoscaling gke-dev-default-pool-efca0068-grp \
    --zone europe-west1-b