不得不说千万不要随意更改版本，我用的1.13的版本，然后学到这一步时，还因yaml文件不同，卡住了很久，然后各种google才找到解决办法
https://www.linuxea.com/2112.html
以前是用heapster来收集资源指标才能看，现在heapster要废弃了。
从k8s v1.8开始后，引入了新的功能，即把资源指标引入api。
资源指标：metrics-server
自定义指标： prometheus,k8s-prometheus-adapter
因此，新一代架构：
1） 核心指标流水线：由kubelet、metrics-server以及由API server提供的api组成；cpu累计利用率、内存实时利用率、pod的资源占用率及容器的磁盘占用率
2） 监控流水线：用于从系统收集各种指标数据并提供终端用户、存储系统以及HPA，他们包含核心指标以及许多非核心指标。非核心指标不能被k8s所解析。
metrics-server是个api server，仅仅收集cpu利用率、内存利用率等。
[root@master ~]# kubectl api-versions
admissionregistration.k8s.io/v1beta1
apiextensions.k8s.io/v1beta1
apiregistration.k8s.io/v1
apiregistration.k8s.io/v1beta1
apps/v1
apps/v1beta1
apps/v1beta2
authentication.k8s.io/v1
authentication.k8s.io/v1beta1
authorization.k8s.io/v1
访问 https://github.com/kubernetes/kubernetes/tree/master/cluster/addons/metrics-server 获取yaml文件，但这个里面的yaml文件更新了。和视频内的有差别
贴出我修改后的yaml文件，留作备用
[root@master metrics-server]# cat auth-delegator.yaml apiVersion: rbac.authorization.k8s.io/v1 kind: ClusterRoleBinding metadata: name: metrics-server:system:auth-delegator labels: kubernetes.io/cluster-service: "true" addonmanager.kubernetes.io/mode: Reconcile roleRef: apiGroup: rbac.authorization.k8s.io kind: ClusterRole name: system:auth-delegator subjects: - kind: ServiceAccount name: metrics-server namespace: kube-system
[root@master metrics-server]# cat auth-reader.yaml apiVersion: rbac.authorization.k8s.io/v1 kind: RoleBinding metadata: name: metrics-server-auth-reader namespace: kube-system labels: kubernetes.io/cluster-service: "true" addonmanager.kubernetes.io/mode: Reconcile roleRef: apiGroup: rbac.authorization.k8s.io kind: Role name: extension-apiserver-authentication-reader subjects: - kind: ServiceAccount name: metrics-server namespace: kube-system
[root@master metrics-server]# cat metrics-apiservice.yaml apiVersion: apiregistration.k8s.io/v1beta1 kind: APIService metadata: name: v1beta1.metrics.k8s.io labels: kubernetes.io/cluster-service: "true" addonmanager.kubernetes.io/mode: Reconcile spec: service: name: metrics-server namespace: kube-system group: metrics.k8s.io version: v1beta1 insecureSkipTLSVerify: true groupPriorityMinimum: 100 versionPriority: 100
关键是这个文件
[root@master metrics-server]# cat metrics-server-deployment.yaml apiVersion: v1 kind: ServiceAccount metadata: name: metrics-server namespace: kube-system labels: kubernetes.io/cluster-service: "true" addonmanager.kubernetes.io/mode: Reconcile --- apiVersion: v1 kind: ConfigMap metadata: name: metrics-server-config namespace: kube-system labels: kubernetes.io/cluster-service: "true" addonmanager.kubernetes.io/mode: EnsureExists data: NannyConfiguration: |- apiVersion: nannyconfig/v1alpha1 kind: NannyConfiguration --- apiVersion: apps/v1 kind: Deployment metadata: name: metrics-server-v0.3.1 namespace: kube-system labels: k8s-app: metrics-server kubernetes.io/cluster-service: "true" addonmanager.kubernetes.io/mode: Reconcile version: v0.3.1 spec: selector: matchLabels: k8s-app: metrics-server version: v0.3.1 template: metadata: name: metrics-server labels: k8s-app: metrics-server version: v0.3.1 annotations: scheduler.alpha.kubernetes.io/critical-pod: '' seccomp.security.alpha.kubernetes.io/pod: 'docker/default' spec: priorityClassName: system-cluster-critical serviceAccountName: metrics-server containers: - name: metrics-server image: mirrorgooglecontainers/metrics-server-amd64:v0.3.1 command: - /metrics-server - --metric-resolution=30s - --kubelet-insecure-tls - --kubelet-preferred-address-types=InternalIP,Hostname,InternalDNS,ExternalDNS,ExternalIP # These are needed for GKE, which doesn't support secure communication yet. # Remove these lines for non-GKE clusters, and when GKE supports token-based auth. #- --kubelet-port=10250 #- --deprecated-kubelet-completely-insecure=true ports: - containerPort: 443 name: https protocol: TCP - name: metrics-server-nanny image: mirrorgooglecontainers/addon-resizer:1.8.4 resources: limits: cpu: 100m memory: 300Mi requests: cpu: 5m memory: 50Mi env: - name: MY_POD_NAME valueFrom: fieldRef: fieldPath: metadata.name - name: MY_POD_NAMESPACE valueFrom: fieldRef: fieldPath: metadata.namespace volumeMounts: - name: metrics-server-config-volume mountPath: /etc/config command: - /pod_nanny - --config-dir=/etc/config - --cpu=100m - --extra-cpu=0.5m - --memory=100Mi - --extra-memory=50Mi - --threshold=5 - --deployment=metrics-server-v0.3.1 - --container=metrics-server - --poll-period=300000 - --estimator=exponential # Specifies the smallest cluster (defined in number of nodes) # # resources will be scaled to. - --minClusterSize=10 volumes: - name: metrics-server-config-volume configMap: name: metrics-server-config tolerations: - key: "CriticalAddonsOnly" operator: "Exists"
[root@master metrics-server]# cat metrics-server-service.yaml apiVersion: v1 kind: Service metadata: name: metrics-server namespace: kube-system labels: addonmanager.kubernetes.io/mode: Reconcile kubernetes.io/cluster-service: "true" kubernetes.io/name: "Metrics-server" spec: selector: k8s-app: metrics-server ports: - port: 443 protocol: TCP targetPort: https
[root@master metrics-server]# cat metrics-server-service.yaml apiVersion: v1 kind: Service metadata: name: metrics-server namespace: kube-system labels: addonmanager.kubernetes.io/mode: Reconcile kubernetes.io/cluster-service: "true" kubernetes.io/name: "Metrics-server" spec: selector: k8s-app: metrics-server ports: - port: 443 protocol: TCP targetPort: https [root@master metrics-server]# cat resource-reader.yaml apiVersion: rbac.authorization.k8s.io/v1 kind: ClusterRole metadata: name: system:metrics-server labels: kubernetes.io/cluster-service: "true" addonmanager.kubernetes.io/mode: Reconcile rules: - apiGroups: - "" resources: - pods - nodes - namespaces - nodes/stats verbs: - get - list - watch - apiGroups: - "extensions" resources: - deployments verbs: - get - list - update - watch --- apiVersion: rbac.authorization.k8s.io/v1 kind: ClusterRoleBinding metadata: name: system:metrics-server labels: kubernetes.io/cluster-service: "true" addonmanager.kubernetes.io/mode: Reconcile roleRef: apiGroup: rbac.authorization.k8s.io kind: ClusterRole name: system:metrics-server subjects: - kind: ServiceAccount name: metrics-server namespace: kube-system
如果从github上下载以上文件apply出错，就用上面的metrics-server-deployment.yaml文件，删掉重新apply一下就可以了
[root@master metrics-server]# kubectl apply -f ./
[root@master ~]# kubectl proxy --port=8080
确保metrics-server-v0.3.1-76b796b-4xgvp是running状态，我当时出现了Error发现是yaml里面有问题，最后该掉running了，该来该去该到上面的最终版
[root@master metrics-server]# kubectl get pods -n kube-system
NAME READY STATUS RESTARTS AGE
canal-mgbc2 3/3 Running 12 3d23h
canal-s4xgb 3/3 Running 23 3d23h
canal-z98bc 3/3 Running 15 3d23h
coredns-78d4cf999f-5shdq 1/1 Running 0 6m4s
coredns-78d4cf999f-xj5pj 1/1 Running 0 5m53s
etcd-master 1/1 Running 13 17d
kube-apiserver-master 1/1 Running 13 17d
kube-controller-manager-master 1/1 Running 19 17d
kube-flannel-ds-amd64-8xkfn 1/1 Running 0
kube-flannel-ds-amd64-t7jpc 1/1 Running 0
kube-flannel-ds-amd64-vlbjz 1/1 Running 0
kube-proxy-ggcbf 1/1 Running 11 17d
kube-proxy-jxksd 1/1 Running 11 17d
kube-proxy-nkkpc 1/1 Running 12 17d
kube-scheduler-master 1/1 Running 19 17d
kubernetes-dashboard-76479d66bb-zr4dd 1/1 Running 0
metrics-server-v0.3.1-76b796b-4xgvp 2/2 Running 0 9s
查看出错日志 -c指定容器名，该pod内有两个容器，metrcis-server只是其中一个，另一个查询方法一样，把名字改掉即可
[root@master metrics-server]# kubectl logs metrics-server-v0.3.1-76b796b-4xgvp -c metrics-server -n kube-system
大致出错的日志内容如下几条；
403 Forbidden", response: "Forbidden (user=system:anonymous, verb=get, resource=nodes, subresource=stats)
E0903 1 manager.go:102] unable to fully collect metrics: [unable to fully scrape metrics from source kubelet_summary:: unable to fetch metrics from Kubelet (): Get https://:10250/stats/summary/: dial tcp: lookup on 10.96.0.10:53: no such host
no response from https://10.101.248.96:443: Get https://10.101.248.96:443: Proxy Error ( Connection refused )
E1109 09:54:49.509521 1 manager.go:102] unable to fully collect metrics: [unable to fully scrape metrics from source kubelet_summary:linuxea.node-2.com: unable to fetch metrics from Kubelet linuxea.node-2.com (10.10.240.203): Get https://10.10.240.203:10255/stats/summary/: dial tcp 10.10.240.203:10255: connect: connection refused, unable to fully scrape metrics from source kubelet_summary:linuxea.node-3.com: unable to fetch metrics from Kubelet linuxea.node-3.com (10.10.240.143): Get https://10.10.240.143:10255/stats/summary/: dial tcp 10.10.240.143:10255: connect: connection refused, unable to fully scrape metrics from source kubelet_summary:linuxea.node-4.com: unable to fetch metrics from Kubelet linuxea.node-4.com (10.10.240.142): Get https://10.10.240.142:10255/stats/summary/: dial tcp 10.10.240.142:10255: connect: connection refused, unable to fully scrape metrics from source kubelet_summary:linuxea.master-1.com: unable to fetch metrics from Kubelet linuxea.master-1.com (10.10.240.161): Get https://10.10.240.161:10255/stats/summary/: dial tcp 10.10.240.161:10255: connect: connection refused, unable to fully scrape metrics from source kubelet_summary:linuxea.node-1.com: unable to fetch metrics from Kubelet linuxea.node-1.com (10.10.240.202): Get https://10.10.240.202:10255/stats/summary/: dial tcp 10.10.240.202:10255: connect: connection refused]
当时我按照网上的方法尝试修改coredns配置，结果搞的日志出现获取所有pod都unable，如下，然后又取消掉了修改，删掉了coredns，让他自己重新生成了俩新的coredns容器
- --kubelet-insecure-tls这种方式是禁用tls验证，一般不建议在生产环境中使用。并且由于DNS是无法解析到这些主机名，使用- --kubelet-preferred-address-types=InternalIP,Hostname,InternalDNS,ExternalDNS,ExternalIP进行规避。还有另外一种方法，修改coredns，不过，我并不建议这样做。
参考这篇：https://github.com/kubernetes-incubator/metrics-server/issues/131
metrics-server unable to fetch pdo metrics for pod
以上为遇到的问题，反正用我上面的yaml绝对保证解决以上所有问题。还有那个flannel改了directrouting之后为啥每次重启集群机器，他就失效呢，我不得不在删掉flannel然后重新生成，这个问题前面文章写到了。
此时执行如下命令就都成功了，item里也有值了
[root@master ~]# curl http://localhost:8080/apis/metrics.k8s.io/v1beta1
{
"kind": "APIResourceList",
"apiVersion": "v1",
"groupVersion": "metrics.k8s.io/v1beta1",
"resources": [
{
"name": "nodes",
"singularName": "",
"namespaced": false,
"kind": "NodeMetrics",
"verbs": [
"get",
"list"
]
},
{
"name": "pods",
"singularName": "",
"namespaced": true,
"kind": "PodMetrics",
"verbs": [
"get",
"list"
]
}
]
[root@master metrics-server]# curl http://localhost:8080/apis/metrics.k8s.io/v1beta1/pods | more
% Total % Received % Xferd Average Speed Time Time Time Current
Dload Upload Total Spent Left Speed
100 14868 0 14868 0 0 1521k 0 --:--:-- --:--:-- --:--:-- 1613k
{
"kind": "PodMetricsList",
"apiVersion": "metrics.k8s.io/v1beta1",
"metadata": {
"selfLink": "/apis/metrics.k8s.io/v1beta1/pods"
},
"items": [
{
"metadata": {
"name": "pod1",
"namespace": "prod",
"selfLink": "/apis/metrics.k8s.io/v1beta1/namespaces/prod/pods/pod1",
"creationTimestamp": "2019-01-29T02:39:12Z"
},
[root@master metrics-server]# kubectl top pods
NAME CPU(cores) MEMORY(bytes)
filebeat-ds-4llpp 1m 2Mi
filebeat-ds-dv49l 1m 5Mi
myapp-0 0m 1Mi
myapp-1 0m 2Mi
myapp-2 0m 1Mi
myapp-3 0m 1Mi
myapp-4 0m 2Mi
[root@master metrics-server]# kubectl top nodes
NAME CPU(cores) CPU% MEMORY(bytes) MEMORY%
master 206m 5% 1377Mi 72%
node1 88m 8% 534Mi 28%
node2 78m 7% 935Mi 49%
自定义指标（prometheus)
大家看到，我们的metrics已经可以正常工作了。不过，metrics只能监控cpu和内存，对于其他指标如用户自定义的监控指标，metrics就无法监控到了。这时就需要另外一个组件叫prometheus。
prometheus的部署非常麻烦。
node_exporter是agent;
PromQL相当于sql语句来查询数据;
k8s-prometheus-adapter:prometheus是不能直接解析k8s的指标的，需要借助k8s-prometheus-adapter转换成api
kube-state-metrics是用来整合数据的。
下面开始部署。
访问 https://github.com/ikubernetes/k8s-prom
[root@master pro]# git clone https://github.com/iKubernetes/k8s-prom.git
先创建一个叫prom的名称空间：
[root@master k8s-prom]# kubectl apply -f namespace.yaml
namespace/prom created
部署node_exporter：
[root@master k8s-prom]# cd node_exporter/
[root@master node_exporter]# ls
node-exporter-ds.yaml node-exporter-svc.yaml
[root@master node_exporter]# kubectl apply -f .
daemonset.apps/prometheus-node-exporter created
service/prometheus-node-exporter created
[root@master node_exporter]# kubectl get pods -n prom
NAME READY STATUS RESTARTS AGE
prometheus-node-exporter-dmmjj 1/1 Running 0 7m
prometheus-node-exporter-ghz2l 1/1 Running 0 7m
prometheus-node-exporter-zt2lw 1/1 Running 0 7m
部署prometheus:
[root@master k8s-prom]# cd prometheus/
[root@master prometheus]# ls
prometheus-cfg.yaml prometheus-deploy.yaml prometheus-rbac.yaml prometheus-svc.yaml
[root@master prometheus]# kubectl apply -f .
configmap/prometheus-config created
deployment.apps/prometheus-server created
clusterrole.rbac.authorization.k8s.io/prometheus created
serviceaccount/prometheus created
clusterrolebinding.rbac.authorization.k8s.io/prometheus created
service/prometheus created
看prom名称空间中的所有资源： pod/prometheus-server-76dc8df7b-hw8xc 处于 Pending 状态，日志显示内存不足
[root@master prometheus]# kubectl logs prometheus-server-556b8896d6-dfqkp -n prom
Warning FailedScheduling 2m52s (x2 over 2m52s) default-scheduler 0/3 nodes are available: 3 Insufficient memory.
修改prometheus-deploy.yaml，删掉内存那三行
resources:
limits:
memory: 2Gi
重新apply
[root@master prometheus]# kubectl apply -f prometheus-deploy.yaml
[root@master prometheus]# kubectl get all -n prom
NAME READY STATUS RESTARTS AGE
pod/prometheus-node-exporter-dmmjj 1/1 Running 0 10m
pod/prometheus-node-exporter-ghz2l 1/1 Running 0 10m
pod/prometheus-node-exporter-zt2lw 1/1 Running 0 10m
pod/prometheus-server-65f5d59585-6l8m8 1/1 Running 0 55s
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
service/prometheus NodePort 10.111.127.64 9090:30090/TCP 56s
service/prometheus-node-exporter ClusterIP None 9100/TCP 10m
NAME DESIRED CURRENT READY UP-TO-DATE AVAILABLE NODE SELECTOR AGE
daemonset.apps/prometheus-node-exporter 3 3 3 3 3 10m
NAME DESIRED CURRENT UP-TO-DATE AVAILABLE AGE
deployment.apps/prometheus-server 1 1 1 1 56s
NAME DESIRED CURRENT READY AGE
replicaset.apps/prometheus-server-65f5d59585 1 1 1 56s
上面我们看到通过NodePorts的方式，可以通过宿主机的30090端口，来访问prometheus容器里面的应用。
最好挂载个pvc的存储，要不这些监控数据过一会就没了。
部署kube-state-metrics，用来整合数据：
[root@master k8s-prom]# cd kube-state-metrics/
[root@master kube-state-metrics]# ls
kube-state-metrics-deploy.yaml kube-state-metrics-rbac.yaml kube-state-metrics-svc.yaml
[root@master kube-state-metrics]# kubectl apply -f .
deployment.apps/kube-state-metrics created
serviceaccount/kube-state-metrics created
clusterrole.rbac.authorization.k8s.io/kube-state-metrics created
clusterrolebinding.rbac.authorization.k8s.io/kube-state-metrics created
service/kube-state-metrics created
[root@master kube-state-metrics]# kubectl get all -n prom
NAME READY STATUS RESTARTS AGE
pod/kube-state-metrics-58dffdf67d-v9klh 1/1 Running 0 14m
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
service/kube-state-metrics ClusterIP 10.111.41.139 8080/TCP 14m
部署k8s-prometheus-adapter,这个需要自制证书：
[root@master k8s-prometheus-adapter]# cd /etc/kubernetes/pki/
[root@master pki]# (umask 077; openssl genrsa -out serving.key 2048)
Generating RSA private key, 2048 bit long modulus
...........................................................................................+++
...............+++
e is 65537 (0x10001)
证书请求：
[root@master pki]# openssl req -new -key serving.key -out serving.csr -subj "/CN=serving"
开始签证：
[root@master pki]# openssl x509 -req -in serving.csr -CA ./ca.crt -CAkey ./ca.key -CAcreateserial -out serving.crt -days 3650
Signature ok
subject=/CN=serving
Getting CA Private Key
创建加密的配置文件：
[root@master pki]# kubectl create secret generic cm-adapter-serving-certs --from-file=serving.crt=./serving.crt --from-file=serving.key=./serving.key -n prom
secret/cm-adapter-serving-certs created
注：cm-adapter-serving-certs是custom-metrics-apiserver-deployment.yaml文件里面的名字。
[root@master pki]# kubectl get secrets -n prom
NAME TYPE DATA AGE
cm-adapter-serving-certs Opaque 2 51s
default-token-knsbg kubernetes.io/service-account-token 3 4h
kube-state-metrics-token-sccdf kubernetes.io/service-account-token 3 3h
prometheus-token-nqzbz kubernetes.io/service-account-token 3 3h
部署k8s-prometheus-adapter：
[root@master k8s-prom]# cd k8s-prometheus-adapter/
[root@master k8s-prometheus-adapter]# ls
custom-metrics-apiserver-auth-delegator-cluster-role-binding.yaml custom-metrics-apiserver-service.yaml
custom-metrics-apiserver-auth-reader-role-binding.yaml custom-metrics-apiservice.yaml
custom-metrics-apiserver-deployment.yaml custom-metrics-cluster-role.yaml
custom-metrics-apiserver-resource-reader-cluster-role-binding.yaml custom-metrics-resource-reader-cluster-role.yaml
custom-metrics-apiserver-service-account.yaml hpa-custom-metrics-cluster-role-binding.yaml
由于k8s v1.11.2和k8s-prometheus-adapter最新版不兼容，1.13的也不兼容，解决办法就是访问https://github.com/DirectXMan12/k8s-prometheus-adapter/tree/master/deploy/manifests下载最新版的custom-metrics-apiserver-deployment.yaml文件，并把里面的namespace的名字改成prom；同时还要下载custom-metrics-config-map.yaml文件到本地来，并把里面的namespace的名字改成prom。
[root@master k8s-prometheus-adapter]# kubectl apply -f .
clusterrolebinding.rbac.authorization.k8s.io/custom-metrics:system:auth-delegator created
rolebinding.rbac.authorization.k8s.io/custom-metrics-auth-reader created
deployment.apps/custom-metrics-apiserver created
clusterrolebinding.rbac.authorization.k8s.io/custom-metrics-resource-reader created
serviceaccount/custom-metrics-apiserver created
service/custom-metrics-apiserver created
apiservice.apiregistration.k8s.io/v1beta1.custom.metrics.k8s.io created
clusterrole.rbac.authorization.k8s.io/custom-metrics-server-resources created
clusterrole.rbac.authorization.k8s.io/custom-metrics-resource-reader created
clusterrolebinding.rbac.authorization.k8s.io/hpa-controller-custom-metrics created
[root@master k8s-prometheus-adapter]# kubectl get all -n prom
NAME READY STATUS RESTARTS AGE
pod/custom-metrics-apiserver-65f545496-64lsz 1/1 Running 0 6m
pod/kube-state-metrics-58dffdf67d-v9klh 1/1 Running 0 4h
pod/prometheus-node-exporter-dmmjj 1/1 Running 0 4h
pod/prometheus-node-exporter-ghz2l 1/1 Running 0 4h
pod/prometheus-node-exporter-zt2lw 1/1 Running 0 4h
pod/prometheus-server-65f5d59585-6l8m8 1/1 Running 0 4h
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
service/custom-metrics-apiserver ClusterIP 10.103.87.246 443/TCP 36m
service/kube-state-metrics ClusterIP 10.111.41.139 8080/TCP 4h
service/prometheus NodePort 10.111.127.64 9090:30090/TCP 4h
service/prometheus-node-exporter ClusterIP None 9100/TCP 4h
NAME DESIRED CURRENT READY UP-TO-DATE AVAILABLE NODE SELECTOR AGE
daemonset.apps/prometheus-node-exporter 3 3 3 3 3 4h
NAME DESIRED CURRENT UP-TO-DATE AVAILABLE AGE
deployment.apps/custom-metrics-apiserver 1 1 1 1 36m
deployment.apps/kube-state-metrics 1 1 1 1 4h
deployment.apps/prometheus-server 1 1 1 1 4h
NAME DESIRED CURRENT READY AGE
replicaset.apps/custom-metrics-apiserver-5f6b4d857d 0 0 0 36m
replicaset.apps/custom-metrics-apiserver-65f545496 1 1 1 6m
replicaset.apps/custom-metrics-apiserver-86ccf774d5 0 0 0 17m
replicaset.apps/kube-state-metrics-58dffdf67d 1 1 1 4h
replicaset.apps/prometheus-server-65f5d59585 1 1 1 4h
最终看到prom名称空间里面的所有资源都是running状态了。
[root@master k8s-prometheus-adapter]# kubectl api-versions
custom.metrics.k8s.io/v1beta1
可以看到custom.metrics.k8s.io/v1beta1这个api了。我那没看到上面这个东西，但是不影响使用
开个代理：
[root@master k8s-prometheus-adapter]# kubectl proxy --port=8080
可以看到指标数据了：
[root@master pki]# curl http://localhost:8080/apis/custom.metrics.k8s.io/v1beta1/
{
"name": "pods/ceph_rocksdb_submit_transaction_sync",
"singularName": "",
"namespaced": true,
"kind": "MetricValueList",
"verbs": [
"get"
]
},
{
"name": "jobs.batch/kube_deployment_created",
"singularName": "",
"namespaced": true,
"kind": "MetricValueList",
"verbs": [
"get"
]
},
{
"name": "jobs.batch/kube_pod_owner",
"singularName": "",
"namespaced": true,
"kind": "MetricValueList",
"verbs": [
"get"
]
},
下面我们就可以愉快的创建HPA了（水平Pod自动伸缩）。
另外，prometheus还可以和grafana整合。如下步骤。
先下载文件grafana.yaml，访问https://github.com/kubernetes/heapster/blob/master/deploy/kube-config/influxdb/grafana.yaml
[root@master pro]# wget https://raw.githubusercontent.com/kubernetes-retired/heapster/master/deploy/kube-config/influxdb/grafana.yaml
修改grafana.yaml文件内容：
把namespace: kube-system改成prom，有两处；
把env里面的下面两个注释掉：
- name: INFLUXDB_HOST
value: monitoring-influxdb
在最有一行加个type: NodePort
ports:
- port: 80
targetPort: 3000
selector:
k8s-app: grafana
type: NodePort
[root@master pro]# kubectl apply -f grafana.yaml
deployment.extensions/monitoring-grafana created
service/monitoring-grafana created
[root@master pro]# kubectl get pods -n prom
NAME READY STATUS RESTARTS AGE
monitoring-grafana-ffb4d59bd-gdbsk 1/1 Running 0 5s
如果还有问题就删掉上面的那几个，重新在apply一下
看到grafana这个pod运行起来了。
[root@master pro]# kubectl get svc -n prom
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
monitoring-grafana NodePort 10.106.164.205 80:32659/TCP 19m
我们可以访问宿主机master ip： http://172.16.1.100:32659
上图端口号是9090，根据自己svc实际端口去填写。除了把80 改成9090.其余不变，为什么是上面的格式，因为他们都处于一个名称空间内，可以通过服务名访问到的。
[root@master pro]# kubectl get svc -n prom
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
custom-metrics-apiserver ClusterIP 10.109.58.249 443/TCP 52m
kube-state-metrics ClusterIP 10.103.52.45 8080/TCP 69m
monitoring-grafana NodePort 10.110.240.31 80:31128/TCP 17m
prometheus NodePort 10.110.19.171 9090:30090/TCP 145m
prometheus-node-exporter ClusterIP None 9100/TCP 146m
然后，就能从界面上看到相应的数据了。
登录下面的网站下载个grafana监控k8s-prometheus的模板： https://grafana.com/dashboards/6417
然后再grafana的界面中导入上面下载的模板：
导入模板之后，就能看到监控数据了：
HPA的没去实际操作，因为以前自己做过了，就不做了，直接复制过来，如有问题自己单独解决
HPA（水平pod自动扩展）
当pod压力大了，会根据负载自动扩展Pod个数以均匀压力。
目前，HPA只支持两个版本，v1版本只支持核心指标的定义（只能根据cpu利用率的指标进行pod的扩展）；
[root@master pro]# kubectl explain hpa.spec.scaleTargetRef
scaleTargetRef：表示基于什么指标来计算pod伸缩的标准
[root@master pro]# kubectl api-versions |grep auto
autoscaling/v1
autoscaling/v2beta1
上面看到分别支持hpav1和hpav2。
下面我们用命令行的方式重新创建一个带有资源限制的pod myapp：
[root@master ~]# kubectl run myapp --image=ikubernetes/myapp:v1 --replicas=1 --requests='cpu=50m,memory=256Mi' --limits='cpu=50m,memory=256Mi' --labels='app=myapp' --expose --port=80
service/myapp created
deployment.apps/myapp created
[root@master ~]# kubectl get pods
NAME READY STATUS RESTARTS AGE
myapp-6985749785-fcvwn 1/1 Running 0 58s
下面我们让myapp 这个pod能自动水平扩展，用kubectl autoscale，其实就是指明HPA控制器的。
[root@master ~]# kubectl autoscale deployment myapp --min=1 --max=8 --cpu-percent=60
horizontalpodautoscaler.autoscaling/myapp autoscaled
--min：表示最小扩展pod的个数
--max：表示最多扩展pod的个数
--cpu-percent：cpu利用率
[root@master ~]# kubectl get hpa
NAME REFERENCE TARGETS MINPODS MAXPODS REPLICAS AGE
myapp Deployment/myapp 0%/60% 1 8 1 4m
[root@master ~]# kubectl get svc
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
myapp ClusterIP 10.105.235.197 80/TCP 19
下面我们把service改成NodePort的方式：
[root@master ~]# kubectl patch svc myapp -p '{"spec":{"type": "NodePort"}}'
service/myapp patched
[root@master ~]# kubectl get svc
NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE
myapp NodePort 10.105.235.197 80:31990/TCP 22m
[root@master ~]# yum install httpd-tools #主要是为了安装ab压测工具
[root@master ~]# kubectl get pods -o wide
NAME READY STATUS RESTARTS AGE IP NODE
myapp-6985749785-fcvwn 1/1 Running 0 25m 10.244.2.84 node2
开始用ab工具压测
[root@master ~]# ab -c 1000 -n 5000000 http://172.16.1.100:31990/index.html
This is ApacheBench, Version 2.3 <$Revision: 1430300 $>
Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
Licensed to The Apache Software Foundation, http://www.apache.org/
Benchmarking 172.16.1.100 (be patient)
多等一会，会看到pods的cpu利用率为98%，需要扩展为2个pod了：
[root@master ~]# kubectl describe hpa
resource cpu on pods (as a percentage of request): 98% (49m) / 60%
Deployment pods: 1 current / 2 desired
[root@master ~]# kubectl top pods
NAME CPU(cores) MEMORY(bytes)
myapp-6985749785-fcvwn 49m （我们设置的总cpu是50m） 3Mi
[root@master ~]# kubectl get pods -o wide
NAME READY STATUS RESTARTS AGE IP NODE
myapp-6985749785-fcvwn 1/1 Running 0 32m 10.244.2.84 node2
myapp-6985749785-sr4qv 1/1 Running 0 2m 10.244.1.105 node1
上面我们看到已经自动扩展为2个pod了，再等一会，随着cpu压力的上升，还会看到自动扩展为4个或更多的pod：
[root@master ~]# kubectl get pods -o wide
NAME READY STATUS RESTARTS AGE IP NODE
myapp-6985749785-2mjrd 1/1 Running 0 1m 10.244.1.107 node1
myapp-6985749785-bgz6p 1/1 Running 0 1m 10.244.1.108 node1
myapp-6985749785-fcvwn 1/1 Running 0 35m 10.244.2.84 node2
myapp-6985749785-sr4qv 1/1 Running 0 5m 10.244.1.105 node1
等压测一停止，pod个数还会收缩为正常个数的。
上面我们用的是hpav1来做的水平pod自动扩展的功能，我们前面也说过，hpa v1版本只能根据cpu利用率括水平自动扩展pod。
下面我们介绍一下hpa v2的功能，它可以根据自定义指标利用率来水平扩展pod。
在使用hpa v2版本前，我们先把前面创建的hpa v1版本删除了，以免和我们测试的hpa v2版本冲突：
[root@master hpa]# kubectl delete hpa myapp
horizontalpodautoscaler.autoscaling "myapp" deleted
好了，下面我们创建一个hpa v2：
[root@master hpa]# cat hpa-v2-demo.yaml
apiVersion: autoscaling/v2beta1 #从这可以看出是hpa v2版本
kind: HorizontalPodAutoscaler
metadata:
name: myapp-hpa-v2
spec:
scaleTargetRef: #根据什么指标来做评估压力
apiVersion: apps/v1 #对谁来做自动扩展
kind: Deployment
name: myapp
minReplicas: 1 #最少副本数量
maxReplicas: 10
metrics: #表示依据哪些指标来进行评估
- type: Resource #表示基于资源进行评估
resource:
name: cpu
targetAverageUtilization: 55 #表示pod cpu使用率超过55%，就自动水平扩展pod个数
- type: Resource
resource:
name: memory #我们知道hpa v1版本只能根据cpu来进行评估，而到了我们的hpa v2版本就可以根据内存来进行评估了
targetAverageValue: 50Mi #表示pod内存使用超过50M，就自动水平扩展pod个数
[root@master hpa]# kubectl apply -f hpa-v2-demo.yaml
horizontalpodautoscaler.autoscaling/myapp-hpa-v2 created
[root@master hpa]# kubectl get hpa
NAME REFERENCE TARGETS MINPODS MAXPODS REPLICAS AGE
myapp-hpa-v2 Deployment/myapp 3723264/50Mi, 0%/55% 1 10 1 37s
我们看到现在只有一个pod
[root@master hpa]# kubectl get pods -o wide
NAME READY STATUS RESTARTS AGE IP NODE
myapp-6985749785-fcvwn 1/1 Running 0 57m 10.244.2.84 node2
开始压测：
[root@master ~]# ab -c 100 -n 5000000 http://172.16.1.100:31990/index.html
看hpa v2的检测情况：
[root@master hpa]# kubectl describe hpa
Metrics: ( current / target )
resource memory on pods: 3756032 / 50Mi
resource cpu on pods (as a percentage of request): 82% (41m) / 55%
Min replicas: 1
Max replicas: 10
Deployment pods: 1 current / 2 desired
[root@master hpa]# kubectl get pods -o wide
NAME READY STATUS RESTARTS AGE IP NODE
myapp-6985749785-8frq4 1/1 Running 0 1m 10.244.1.109 node1
myapp-6985749785-fcvwn 1/1 Running 0 1h 10.244.2.84 node2
看到自动扩展出了2个Pod。等压测一停止，pod个数还会收缩为正常个数的。
将来我们不光可以用hpa v2，根据cpu和内存使用率进行伸缩Pod个数，还可以根据http并发量等。
比如下面的：
[root@master hpa]# cat hpa-v2-custom.yaml
apiVersion: autoscaling/v2beta1 #从这可以看出是hpa v2版本
kind: HorizontalPodAutoscaler
metadata:
name: myapp-hpa-v2
spec:
scaleTargetRef: #根据什么指标来做评估压力
apiVersion: apps/v1 #对谁来做自动扩展
kind: Deployment
name: myapp
minReplicas: 1 #最少副本数量
maxReplicas: 10
metrics: #表示依据哪些指标来进行评估
- type: Pods #表示基于资源进行评估
pods:
metricName: http_requests#自定义的资源指标
targetAverageValue: 800m #m表示个数,表示并发数800
关于并发数的hpa，具体镜像可以参考https://hub.docker.com/r/ikubernetes/metrics-app/