1.1 Prometheus原理

1.1.1 Prometheus架构

Kubernetes监控方案之Grafana + Alertmanager + Prometheus_Alertmanager


数据采集模块:最左边的Prometheus target就是数据采集对象,而Retrieval则负责采集这些数据,并同时支持push,pull两种采集方式。

  Pull模式:有服务端的采集模式来触发,只要采集目标,即target提供了http接口就可以采集,也是最常用的采集方式。后面描述的基本都是采用这种方式采集。

  Push模式:由各个采集目标主动向push gateway推送指标,再由服务器端拉取。

 

存储模块:为了保证监控数据持久化,由TSDB进行存储,是专门为时间序列数据设置的数据库,能够以时间为索引进行存储。

 

数据查询和处理模块:TSDB在提供存储的同时还提供了数据查询和处理的基本功能,就是PromSQL,这也是告警系统以及可视化页面的基础。

 

告警模块:由Alertmanager进行告警,能够动态进行分组,静默,抑制,减少告警数量,能够支持多种告警方式。

 

可视化模块:Grafana实现的多维度可视化界面。


1.1.2 怎么来监控?

  Prometheus通过pushpull两种数据采集方式采集数据,通过在配置文件中定义的相关采集项,对k8s不同组件甚至开源应用如mysqlredis进行数据采集,将数据存储在prometheus server中的TSDB时序数据库中,当rules在指定时间段被触发后推送给Alertmanager进行报警处理,Grafana通过PromQL拉取监控数据进行可视化图表展示。

psnode_exporter采集器相关地址

使用文档:https://prometheus.io/docs/guides/node-exporter/

GitHubhttps://github.com/prometheus/node_exporter

exporter列表:https://prometheus.io/docs/instrumenting/exporters/


1.1.3 Prometheus标识

Prometheus将所有数据存储为时间序列;具有相同度量名称以及标签属于同一个指标。

每个时间序列都由度量标准名称和一组键值对(也成为标签)唯一标识。

时间序列格式:

<metric name>{<label name>=<label value>, ...}

示例如下:

container_cpu_user_seconds_total{beta_kubernetes_io_arch="amd64",beta_kubernetes_io_os="linux",container_name="POD",id="/kubepods/besteffort/pod011eaa38-af5c-11e8-a65f-801844f11504/b659e04fa88e0f3080612d6479dec32d41afe632efd4f43f80714245be899d14",image="k8s.gcr.io/pause-amd64:3.1",instance="h-ldocker-02",job="kubernetes-cadvisor",kubernetes_io_hostname="h-ldocker-02",name="k8s_POD_consul-client-0_plms-pro_011eaa38-af5c-11e8-a65f-801844f11504_0",namespace="plms-pro",pod_name="consul-client-0",rolehost="k8smaster"}


1.2 Prometheus容器化部署及应用

  解压k8s-prometheus-grafana.tar.gz,部署prometheus.svc.yml,rbac-setup.yaml, prometheus-rules.yaml ,configmap.yaml,prometheus.deploy.yaml。

 

prometheus.svc:配置prometheusnodePort30003,供外部访问,调试Promsql

[root@PH-K8S-M01 k8s-prometheus-grafana]# cat prometheus.svc.yml 

    ---

    kind: Service

    apiVersion: v1

    metadata:

      labels:

        app: prometheus

      name: prometheus

      namespace: kube-system

    spec:

      type: NodePort

      ports:

      - port: 9090

        targetPort: 9090

        nodePort: 30003

      selector:

        app: prometheus


rbac-setup:配置prometheus用户在k8s的组件访问权限。

[root@PH-K8S-M01 k8s-prometheus-grafana]# cat rbac-setup.yaml 

apiVersion: rbac.authorization.k8s.io/v1

kind: ClusterRole

metadata:

  name: prometheus

rules:

- apiGroups: [""]

  resources:

  - nodes

  - nodes/proxy

  - services

  - endpoints

  - pods

  verbs: ["get", "list", "watch"]

- apiGroups:

  - extensions

  resources:

  - ingresses

  verbs: ["get", "list", "watch"]

- nonResourceURLs: ["/metrics"]

  verbs: ["get"]

---

apiVersion: v1

kind: ServiceAccount

metadata:

  name: prometheus

  namespace: kube-system

---

apiVersion: rbac.authorization.k8s.io/v1

kind: ClusterRoleBinding

metadata:

  name: prometheus

roleRef:

  apiGroup: rbac.authorization.k8s.io

  kind: ClusterRole

  name: prometheus

subjects:

- kind: ServiceAccount

  name: prometheus

  namespace: kube-system


prometheus-rules:设置prometheus的监控项。分为general.rulesnode.rules,分别监控k8spod以及node节点的数据。

以下图InstanceDown监控项为例,监控的数据来自于”expr: up ==0“,“up“监控的是prometheus target中的组件健康状态,如apiserver,当up==0时表示异常,当异常持续两分钟,即”for: 2m”,就会将异常数据传递给alertmanager进行分组报警处理。

[root@PH-K8S-M01 k8s-prometheus-grafana]# cat prometheus-rules.yaml 

apiVersion: v1

kind: ConfigMap

metadata:

  name: prometheus-rules

  namespace: kube-system

data:

  general.rules: |

    groups:

    - name: general.rules

      rules:

      - alert: InstanceDown

        expr: up == 0

        for: 2m

        labels:

          severity: error 

        annotations:

          summary: "Instance {{ $labels.instance }} 停止工作"

          description: "{{ $labels.instance }} job {{ $labels.job }} 已经停止5分钟以上."

    

      - alert: PodCpuuse

        expr: sum(rate(container_cpu_usage_seconds_total{image!=""}[1m])) by (pod_name, namespace) / (sum(container_spec_cpu_quota{image!=""}/100000) by (pod_name, namespace)) * 100 > 80

        for: 5m

        labels:

          severity: warn 

        annotations:

          summary: "Namespaces: {{ $labels.namespace }} | PodName: {{ $labels.pod_name }} CPU使用率过高"

          description: "Namespaces: {{ $labels.namespace }} | PodName: {{ $labels.pod_name }}  CPU使用率大于80%. (当前值:{{ $value }}%)"


      - alert: PodMemoryuse

        expr: sum(container_memory_rss{image!=""}) by(pod_name, namespace) / sum(container_spec_memory_limit_bytes{image!=""}) by(pod_name, namespace) * 100 != +inf > 80

        for: 5m

        labels:

          severity: warn 

        annotations:

          summary: "Namespaces: {{ $labels.namespace }} | PodName: {{ $labels.pod_name }} 内存使用率过高"

          description: "Namespaces: {{ $labels.namespace }} | PodName: {{ $labels.pod_name }}  内存使用率大于80%. (当前值:{{ $value }}%)"


      - alert: PodFailed

        expr: sum (kube_pod_status_phase{phase="Failed"}) by (pod,namespace) > 0

        for: 1m

        labels:

          severity: error 

        annotations:

          summary: "Namespaces: {{ $labels.namespace }} | PodName: {{ $labels.pod }} pod status is Failed"

          description: "Namespaces: {{ $labels.namespace }} | PodName: {{ $labels.pod }} pod status is Failed. (当前值:{{ $value }})"


      - alert: PodPending

        expr: sum (kube_pod_status_phase{phase="Pending"}) by (pod,namespace) > 0

        for: 1m

        labels:

          severity: error 

        annotations:

          summary: "Namespaces: {{ $labels.namespace }} | PodName: {{ $labels.pod }} pod status is Pending"

          description: "Namespaces: {{ $labels.namespace }} | PodName: {{ $labels.pod }} pod status is Pending. (当前值:{{ $value }})"


      - alert: PodNetworkReceive

        expr: sum (rate (container_network_receive_bytes_total{image!="",name=~"^k8s_.*"}[5m]) /1000) by (pod_name,namespace)  > 20000

        for: 5m

        labels:

          severity: warn

        annotations:

          summary: "Namespaces: {{ $labels.namespace }} | PodName: {{ $labels.pod_name }}  接受到的网络流量过大"

          description: "Namespaces: {{ $labels.namespace }} | PodName: {{ $labels.pod_name }}  接受到的网络流量大于20MB/s. (当前值:{{ $value }}B/s)"


      - alert: PodNetworkTransmit

        expr: sum (rate (container_network_transmit_bytes_total{image!="",name=~"^k8s_.*"}[5m]) /1000) by (pod_name,namespace) > 20000

        for: 5m

        labels:

          severity: warn 

        annotations:

          summary: "Namespaces: {{ $labels.namespace }} | PodName: {{ $labels.pod_name }}  传输的网络流量过大"

          description: "Namespaces: {{ $labels.namespace }} | PodName: {{ $labels.pod_name }}  传输的网络流量大于20MB/s. (当前值:{{ $value }}B/s)"


      - alert: PodRestart

        expr: sum (changes (kube_pod_container_status_restarts_total[1m])) by (pod,namespace) > 0

        for: 1m

        labels:

          severity: warn 

        annotations:

          summary: "Namespaces: {{ $labels.namespace }} | PodName: {{ $labels.pod }} pod is restart"

          description: "Namespaces: {{ $labels.namespace }} | PodName: {{ $labels.pod }}  在一分钟内重启. (当前值:{{ $value }})"


  node.rules: |

    groups:

    - name: node.rules

      rules:

      - alert: NodeFilesystemUsage

        expr: 100 - (node_filesystem_free_bytes{fstype=~"ext4|xfs"} / node_filesystem_size_bytes{fstype=~"ext4|xfs"} * 100) > 80 

        for: 5m

        labels:

          severity: warning 

        annotations:

          summary: "Instance {{ $labels.instance }} : {{ $labels.mountpoint }} 分区使用率过高"

          description: "{{ $labels.instance }}: {{ $labels.mountpoint }} 分区使用大于80% (当前值: {{ $value }})"


      - alert: NodeMemoryUsage

        expr: 100 - (node_memory_MemFree_bytes+node_memory_Cached_bytes+node_memory_Buffers_bytes) / node_memory_MemTotal_bytes * 100 > 80

        for: 5m

        labels:

          severity: warning

        annotations:

          summary: "Instance {{ $labels.instance }} 内存使用率过高"

          description: "{{ $labels.instance }}内存使用大于80% (当前值: {{ $value }})"


      - alert: NodeCPUUsage    

        expr: 100 - (avg(irate(node_cpu_seconds_total{mode="idle"}[5m])) by (instance) * 100) > 60 

        for: 5m

        labels:

          severity: warning

        annotations:

          summary: "Instance {{ $labels.instance }} CPU使用率过高"       

          description: "{{ $labels.instance }}CPU使用大于60% (当前值: {{ $value }})"


configmap.yamlprometheus 配置文件,主要是去采集监控的时序数据,通过cadvisor,获取容器的cpu,内存,网络io等基本数据,通过kubernetes-node获取node主机的相关监控数据,通过kubernetes-services-endpoint来获取kube-state-metrics组件数据,configmap是以角色(role)来定义收集的,比如node,service,pod,endpoints,ingress等。

[root@PH-K8S-M01 k8s-prometheus-grafana]# cat configmap.yaml 

apiVersion: v1

kind: ConfigMap

metadata:

  name: prometheus-config

  namespace: kube-system

data:

  prometheus.yml: |

    rule_files:              ##在配置文件中显性指定prometheus rules文件路径,需要挂载在deployment中;

    - /etc/config/rules/*.rules


    global:

      scrape_interval:     15s  ##prometheus采集数据间隔

      evaluation_interval: 15s


    

    scrape_configs:


    

    - job_name: 'kubernetes-apiservers'

      kubernetes_sd_configs:

      - role: endpoints

      scheme: https

      tls_config:

        ca_file: /var/run/secrets/kubernetes.io/serviceaccount/ca.crt

      bearer_token_file: /var/run/secrets/kubernetes.io/serviceaccount/token

      relabel_configs:

      - source_labels: [__meta_kubernetes_namespace, __meta_kubernetes_service_name, __meta_kubernetes_endpoint_port_name]

        action: keep

        regex: default;kubernetes;https

    

    - job_name: kubernetes-nodes   ##node-exporter采集node数据

      scrape_interval: 30s

      static_configs:

      - targets:

        - *

 


    - job_name: 'kubernetes-nodes-kubelet'

      kubernetes_sd_configs:

      - role: node

      scheme: https

      tls_config:

        ca_file: /var/run/secrets/kubernetes.io/serviceaccount/ca.crt

      bearer_token_file: /var/run/secrets/kubernetes.io/serviceaccount/token

      relabel_configs:

      - action: labelmap

        regex: __meta_kubernetes_node_label_(.+)

      - target_label: __address__

        replacement: kubernetes.default.svc:443

      - source_labels: [__meta_kubernetes_node_name]

        regex: (.+)

        target_label: __metrics_path__

        replacement: /api/v1/nodes/${1}/proxy/metrics


    - job_name: 'kubernetes-nodes-cadvisor'

      kubernetes_sd_configs:

      - role: node

      scheme: https

      tls_config:

        ca_file: /var/run/secrets/kubernetes.io/serviceaccount/ca.crt

      bearer_token_file: /var/run/secrets/kubernetes.io/serviceaccount/token

      relabel_configs:

      - action: labelmap

        regex: __meta_kubernetes_node_label_(.+)

      - target_label: __address__

        replacement: kubernetes.default.svc:443

      - source_labels: [__meta_kubernetes_node_name]

        regex: (.+)

        target_label: __metrics_path__

        replacement: /api/v1/nodes/${1}/proxy/metrics/cadvisor


    - job_name: 'kubernetes-service-endpoints'

      kubernetes_sd_configs:

      - role: endpoints

      relabel_configs:

      - source_labels: [__meta_kubernetes_service_annotation_prometheus_io_scrape]

        action: keep

        regex: true

      - source_labels: [__meta_kubernetes_service_annotation_prometheus_io_scheme]

        action: replace

        target_label: __scheme__

        regex: (https?)

      - source_labels: [__meta_kubernetes_service_annotation_prometheus_io_path]

        action: replace

        target_label: __metrics_path__

        regex: (.+)

      - source_labels: [__address__, __meta_kubernetes_service_annotation_prometheus_io_port]

        action: replace

        target_label: __address__

        regex: ([^:]+)(?::\d+)?;(\d+)

        replacement: $1:$2

      - action: labelmap

        regex: __meta_kubernetes_service_label_(.+)

      - source_labels: [__meta_kubernetes_namespace]

        action: replace

        target_label: kubernetes_namespace

      - source_labels: [__meta_kubernetes_service_name]

        action: replace

        target_label: kubernetes_name


    - job_name: 'kubernetes-services'

      kubernetes_sd_configs:

      - role: service

      metrics_path: /probe

      params:

        module: [http_2xx]

      relabel_configs:

      - source_labels: [__meta_kubernetes_service_annotation_prometheus_io_probe]

        action: keep

        regex: true

      - source_labels: [__address__]

        target_label: __param_target

      - target_label: __address__

        replacement: blackbox-exporter.example.com:9115

      - source_labels: [__param_target]

        target_label: instance

      - action: labelmap

        regex: __meta_kubernetes_service_label_(.+)

      - source_labels: [__meta_kubernetes_namespace]

        target_label: kubernetes_namespace

      - source_labels: [__meta_kubernetes_service_name]

        target_label: kubernetes_name


    - job_name: 'kubernetes-ingresses'

      kubernetes_sd_configs:

      - role: ingress

      relabel_configs:

      - source_labels: [__meta_kubernetes_ingress_annotation_prometheus_io_probe]

        action: keep

        regex: true

      - source_labels: [__meta_kubernetes_ingress_scheme,__address__,__meta_kubernetes_ingress_path]

        regex: (.+);(.+);(.+)

        replacement: ${1}://${2}${3}

        target_label: __param_target

      - target_label: __address__

        replacement: blackbox-exporter.example.com:9115

      - source_labels: [__param_target]

        target_label: instance

      - action: labelmap

        regex: __meta_kubernetes_ingress_label_(.+)

      - source_labels: [__meta_kubernetes_namespace]

        target_label: kubernetes_namespace

      - source_labels: [__meta_kubernetes_ingress_name]

        target_label: kubernetes_name


    - job_name: 'kubernetes-pods'

      kubernetes_sd_configs:

      - role: pod

      relabel_configs:

      - source_labels: [__meta_kubernetes_pod_annotation_prometheus_io_scrape]

        action: keep

        regex: true

      - source_labels: [__meta_kubernetes_pod_annotation_prometheus_io_path]

        action: replace

        target_label: __metrics_path__

        regex: (.+)

      - source_labels: [__address__, __meta_kubernetes_pod_annotation_prometheus_io_port]

        action: replace

        regex: ([^:]+)(?::\d+)?;(\d+)

        replacement: $1:$2

        target_label: __address__

      - action: labelmap

        regex: __meta_kubernetes_pod_label_(.+)

      - source_labels: [__meta_kubernetes_namespace]

        action: replace

        target_label: kubernetes_namespace

      - source_labels: [__meta_kubernetes_pod_name]

        action: replace

        target_label: kubernetes_pod_name

    alerting:                ##alertmanager与prometheus集成配置

      alertmanagers:

      - static_configs:

          - targets: ["alertmanager:80"]


⑤部署prometheus deployment


[root@PH-K8S-M01 k8s-prometheus-grafana]# cat prometheus.deploy.yml 

---

apiVersion: apps/v1beta2

kind: Deployment

metadata:

  labels:

    name: prometheus-deployment

  name: prometheus

  namespace: kube-system

spec:

  replicas: 1

  selector:

    matchLabels:

      app: prometheus

  template:

    metadata:

      labels:

        app: prometheus

    spec:

      containers:

      - image: prom/prometheus:v2.0.0

        name: prometheus

        command:

        - "/bin/prometheus"

        args:

        - "--config.file=/etc/prometheus/prometheus.yml"

        - "--storage.tsdb.path=/prometheus"

        - "--storage.tsdb.retention=24h"

        ports:

        - containerPort: 9090

          protocol: TCP

        volumeMounts:

        - mountPath: "/prometheus"

          name: data

        - mountPath: "/etc/prometheus"

          name: config-volume

        - name: prometheus-rules        ###指定prometheus rule监控项数据挂载,需要在configmap中指定rules路径

          mountPath: /etc/config/rules

        - name: localtime

          mountPath: /etc/localtime

        resources:

          requests:

            cpu: 100m

            memory: 100Mi

          limits:

            cpu: 500m

            memory: 2500Mi

      serviceAccountName: prometheus    

      volumes:

      - name: data

        hostPath:

          path: /var/lib/docker/prometheus 

      - name: config-volume

        configMap:

          name: prometheus-config   

      - name: prometheus-rules

        configMap:

          name: prometheus-rules        

      - name: localtime

        hostPath:

          path: /etc/localtime


至此prometheus部署完成,访问http://IP:30003

查看rules是否正常:

Kubernetes监控方案之Grafana + Alertmanager + Prometheus_Kubernetes_02


查看target是否正常:

Kubernetes监控方案之Grafana + Alertmanager + Prometheus_Prometheus_03


1.3 部署node-exporter

  node-exporter的部署在这里采用二进制包的形式部署,因为发现采用容器化部署node-exporter获取到的数据不是很准确,所以在这里采用二进制的方式,运行一下部署脚本即可,脚本如下,注意,需要在每个node节点执行一下脚本,启动端口为9100

#!/bin/bash


wget https://github.com/prometheus/node_exporter/releases/download/v0.17.0/node_exporter-0.17.0.linux-amd64.tar.gz


tar zxf node_exporter-0.17.0.linux-amd64.tar.gz

mv node_exporter-0.17.0.linux-amd64 /usr/local/node_exporter


cat <<EOF >/usr/lib/systemd/system/node_exporter.service

[Unit]

Description=https://prometheus.io


[Service]

Restart=on-failure

ExecStart=/usr/local/node_exporter/node_exporter --collector.systemd --collector.systemd.unit-whitelist=(docker|kubelet|kube-proxy|flanneld).service


[Install]

WantedBy=multi-user.target

EOF


systemctl daemon-reload

systemctl enable node_exporter

systemctl restart node_exporter 


1.4 部署kube-state-metrics

Kubernetes集群上Pod, DaemonSet, Deployment, Job, CronJob等各种资源对象的状态需要监控,这也反映了使用这些资源部署的应用的状态。但通过查看前面Prometheus从k8s集群拉取的指标(这些指标主要来自apiserver和kubelet中集成的cAdvisor),并没有具体的各种资源对象的状态指标。对于Prometheus来说,当然是需要引入新的exporter来暴露这些指标,Kubernetes提供了一个kube-state-metrics正是我们需要。

部署kube-state-metrics-rbac.yaml进行授权认证

[root@PH-K8S-M01 k8s-prometheus-grafana]# cat kube-state-metrics-rbac.yaml

apiVersion: v1

kind: ServiceAccount

metadata:

  name: kube-state-metrics

  namespace: kube-system

  labels:

    kubernetes.io/cluster-service: "true"

    addonmanager.kubernetes.io/mode: Reconcile

---

apiVersion: rbac.authorization.k8s.io/v1

kind: ClusterRole

metadata:

  name: kube-state-metrics

  labels:

    kubernetes.io/cluster-service: "true"

    addonmanager.kubernetes.io/mode: Reconcile

rules:

- apiGroups: [""]

  resources:

  - configmaps

  - secrets

  - nodes

  - pods

  - services

  - resourcequotas

  - replicationcontrollers

  - limitranges

  - persistentvolumeclaims

  - persistentvolumes

  - namespaces

  - endpoints

  verbs: ["list", "watch"]

- apiGroups: ["extensions"]

  resources:

  - daemonsets

  - deployments

  - replicasets

  verbs: ["list", "watch"]

- apiGroups: ["apps"]

  resources:

  - statefulsets

  verbs: ["list", "watch"]

- apiGroups: ["batch"]

  resources:

  - cronjobs

  - jobs

  verbs: ["list", "watch"]

- apiGroups: ["autoscaling"]

  resources:

  - horizontalpodautoscalers

  verbs: ["list", "watch"]

---

apiVersion: rbac.authorization.k8s.io/v1

kind: Role

metadata:

  name: kube-state-metrics-resizer

  namespace: kube-system

  labels:

    kubernetes.io/cluster-service: "true"

    addonmanager.kubernetes.io/mode: Reconcile

rules:

- apiGroups: [""]

  resources:

  - pods

  verbs: ["get"]

- apiGroups: ["extensions"]

  resources:

  - deployments

  resourceNames: ["kube-state-metrics"]

  verbs: ["get", "update"]

---

apiVersion: rbac.authorization.k8s.io/v1 

kind: ClusterRoleBinding

metadata:

  name: kube-state-metrics

  labels:

    kubernetes.io/cluster-service: "true"

    addonmanager.kubernetes.io/mode: Reconcile

roleRef:

  apiGroup: rbac.authorization.k8s.io

  kind: ClusterRole

  name: kube-state-metrics

subjects:

- kind: ServiceAccount

  name: kube-state-metrics

  namespace: kube-system

---

apiVersion: rbac.authorization.k8s.io/v1

kind: RoleBinding

metadata:

  name: kube-state-metrics

  namespace: kube-system

  labels:

    kubernetes.io/cluster-service: "true"

    addonmanager.kubernetes.io/mode: Reconcile

roleRef:

  apiGroup: rbac.authorization.k8s.io

  kind: Role

  name: kube-state-metrics-resizer

subjects:

- kind: ServiceAccount

  name: kube-state-metrics

  namespace: kube-system

  

部署kube-state-metrics-service.yaml

[root@PH-K8S-M01 k8s-prometheus-grafana]# cat kube-state-metrics-service.yaml

apiVersion: v1

kind: Service

metadata:

  name: kube-state-metrics

  namespace: kube-system

  labels:

    kubernetes.io/cluster-service: "true"

    addonmanager.kubernetes.io/mode: Reconcile

    kubernetes.io/name: "kube-state-metrics"

  annotations:

    prometheus.io/scrape: 'true'

spec:

  ports:

  - name: http-metrics

    port: 8080

    targetPort: http-metrics

    protocol: TCP

  - name: telemetry

    port: 8081

    targetPort: telemetry

    protocol: TCP

  selector:

    k8s-app: kube-state-metrics


部署kube-state-metrics-deployment.yaml

[root@PH-K8S-M01 k8s-prometheus-grafana]# cat kube-state-metrics-deployment.yaml

apiVersion: apps/v1

kind: Deployment

metadata:

  name: kube-state-metrics

  namespace: kube-system

  labels:

    k8s-app: kube-state-metrics

    kubernetes.io/cluster-service: "true"

    addonmanager.kubernetes.io/mode: Reconcile

    version: v1.3.0

spec:

  selector:

    matchLabels:

      k8s-app: kube-state-metrics

      version: v1.3.0

  replicas: 1

  template:

    metadata:

      labels:

        k8s-app: kube-state-metrics

        version: v1.3.0

      annotations:

        scheduler.alpha.kubernetes.io/critical-pod: ''

    spec:

      priorityClassName: system-cluster-critical

      serviceAccountName: kube-state-metrics

      containers:

      - name: kube-state-metrics

        image: **kube-state-metrics:v1.3.0

        ports:

        - name: http-metrics

          containerPort: 8080

        - name: telemetry

          containerPort: 8081

        readinessProbe:

          httpGet:

            path: /healthz

            port: 8080

          initialDelaySeconds: 5

          timeoutSeconds: 5

      - name: addon-resizer

        image: **addon-resizer:1.8.3

        resources:

          limits:

            cpu: 100m

            memory: 30Mi

          requests:

            cpu: 100m

            memory: 30Mi

        env:

          - name: MY_POD_NAME

            valueFrom:

              fieldRef:

                fieldPath: metadata.name

          - name: MY_POD_NAMESPACE

            valueFrom:

              fieldRef:

                fieldPath: metadata.namespace

        volumeMounts:

          - name: config-volume

            mountPath: /etc/config

        command:

          - /pod_nanny

          - --config-dir=/etc/config

          - --container=kube-state-metrics

          - --cpu=100m

          - --extra-cpu=1m

          - --memory=100Mi

          - --extra-memory=2Mi

          - --threshold=5

          - --deployment=kube-state-metrics

      volumes:

        - name: config-volume

          configMap:

            name: kube-state-metrics-config

---

# Config map for resource configuration.

apiVersion: v1

kind: ConfigMap

metadata:

  name: kube-state-metrics-config

  namespace: kube-system

  labels:

    k8s-app: kube-state-metrics

    kubernetes.io/cluster-service: "true"

    addonmanager.kubernetes.io/mode: Reconcile

data:

  NannyConfiguration: |-

    apiVersion: nannyconfig/v1alpha1

    kind: NannyConfiguration


验证:在prometheus上查看target是否正常获取到,Grath页面能够获取到kube相关配置数据,如下

Kubernetes监控方案之Grafana + Alertmanager + Prometheus_Alertmanager_04

Kubernetes监控方案之Grafana + Alertmanager + Prometheus_Prometheus_05


1.5 PromSQL基本语法

Kubernetes监控方案之Grafana + Alertmanager + Prometheus_Kubernetes_06

Kubernetes监控方案之Grafana + Alertmanager + Prometheus_Kubernetes_07


指标在grafana的体现:

查看某个指标的使用率,比如cpu使用率

用到rate函数,rate用来计算两个间隔时间内发生的变化率。如

rate(指标名{筛选条件}[时间间隔])

这两个函数在 promethues 中经常用来计算增量或者速率,在使用时需要指定时间范围如[1m]

irate(): 计算的是给定时间窗口内的每秒瞬时增加速率。

rate(): 计算的是给定时间窗口内的每秒的平均值。

比如查看1分钟内非idle的cpu使用率rate(node_cpu_seconds_total{mode!="idle"}[1m])

Promsql表达式需要在prometheus调试页面调试好再放入grafana进行数据展示,在prometheus-rule中编写的expr作为监控项,其实就是grafana中的各个监控指标,稍加修改,或者直接挪用作为prometheus监控指标。

Node节点监控表达式

CPU使用率:获取5分钟cpu每秒空闲的平均值,从而得出cpu使用率

100-(avg(irate(node_cpu_seconds_total{mode="idle"}[5m]))by(instance)*100)

内存使用率:将内存buffer,cache,free相加得出总的空闲内存,除以总内存大小,从而得出内存使用率

100-(node_memory_MemFree_bytes+node_memory_Cached_bytes+node_memory_Buffers_bytes)/node_memory_MemTotal_bytes*100

磁盘使用率:同理可得

100-(node_filesystem_free_bytes{mountpoint="/",fstype=~"ext4|xfs"}/node_filesystem_size_bytes{mountpoint="/",fstype=~"ext4|xfs"}*100)


容器监控表达式

target组件的健康状态:可以通过添加apiserver,scheduler,controller进行监控,1为正常,0为异常

up == 0

容器cpu使用率:获取1分钟内容器cpu使用值,除以容器cpu limit值,获取到cpu使用率

sum(rate(container_cpu_usage_seconds_total{image!=""}[1m])) by (pod_name, namespace) / (sum(container_spec_cpu_quota{image!=""}/100000) by (pod_name, namespace)) * 100 > 80



容器内存使用率:容器实际申请的内存(RSS)除以限定的内存值,获取到内存使用率

sum(container_memory_rss{image!=""}) by(pod_name, namespace) / sum(container_spec_memory_limit_bytes{image!=""}) by(pod_name, namespace) * 100 != +inf > 80


容器状态为Failed,0为正常,1为Failed

sum (kube_pod_status_phase{phase="Failed"}) by (pod,namespace) > 0


容器状态为Pending,0为正常,1为异常

sum (kube_pod_status_phase{phase="Pending"}) by (pod,namespace) > 0


容器 network 入io,若超过20MB/s则告警(视情况而定)

sum (rate (container_network_receive_bytes_total{image!="",name=~"^k8s_.*"}[5m]) /1000) by (pod_name,namespace)  > 20000


容器 network 出io,若超过20MB/s则告警(视情况而定)

sum (rate (container_network_transmit_bytes_total{image!="",name=~"^k8s_.*"}[5m]) /1000) by (pod_name,namespace) > 20000


容器是否自动重启,如内存溢出等情况

sum (changes (kube_pod_container_status_restarts_total[1m])) by (pod,namespace) > 0


2.1 Grafana容器化部署及应用 

部署grafana-svc.yaml

[root@PH-K8S-M01 k8s-prometheus-grafana]# cat grafana-svc.yaml

apiVersion: v1

kind: Service

metadata:

  name: grafana

  namespace: kube-system

  labels:

    app: grafana

    component: core

spec:

  type: NodePort

  ports:

    - port: 3000

      targetPort: 3000

      nodePort: 30002

  selector:

    app: grafana

component: core


部署grafana-deploy.yaml使用的是5.4.2较新版本

apiVersion: extensions/v1beta1

kind: Deployment

metadata:

  name: grafana-core

  namespace: kube-system

  labels:

    app: grafana

    component: core

spec:

  replicas: 1

  template:

    metadata:

      labels:

        app: grafana

        component: core

    spec:

      containers:

      - image: grafana/grafana:5.4.2

        name: grafana-core

        imagePullPolicy: IfNotPresent

        # env:

        resources:

          # keep request = limit to keep this container in guaranteed class

          limits:

            cpu: 2000m

            memory: 2000Mi

          requests:

            cpu: 1000m

            memory: 2000Mi

        env:

          # The following env variables set up basic auth twith the default admin user and admin password.

          - name: GF_AUTH_BASIC_ENABLED

            value: "true"

          - name: GF_AUTH_ANONYMOUS_ENABLED

            value: "false"

          # - name: GF_AUTH_ANONYMOUS_ORG_ROLE

          #   value: Admin

          # does not really work, because of template variables in exported dashboards:

          # - name: GF_DASHBOARDS_JSON_ENABLED

          #   value: "true"

        readinessProbe:

          httpGet:

            path: /login

            port: 3000

          # initialDelaySeconds: 30

          # timeoutSeconds: 1

        volumeMounts:

        - name: grafana-storage

          mountPath: /var/lib/grafana


      volumes:

      - name: grafana-storage

        hostPath:

          path: /var/lib/grafana

          type: DirectoryOrCreate


2.2 Grafana 配置Prometheus数据源

访问地址:IP: 30002

第一次登陆grafana默认用户名密码均为admin

配置数据源


Kubernetes监控方案之Grafana + Alertmanager + Prometheus_Alertmanager_08

Kubernetes监控方案之Grafana + Alertmanager + Prometheus_Prometheus_09


2.3 Grafana加载自定义的监控模板

Kubernetes监控方案之Grafana + Alertmanager + Prometheus_Prometheus_10


导入已经提前定义好的模板 

Kubernetes AllPod monitoring.json:监控所有pod,包括历史pod

Kubernetes 集群资源监控.json:监控当前在线pod

Kubernetes Node监控.json:监控node节点。


Kubernetes监控方案之Grafana + Alertmanager + Prometheus_Alertmanager_11


也可以输入模板编号,在线获取模板,推荐使用

集群资源监控:3119

资源状态监控:6417

Node监控:9276

 

配置好的grafana页面如下

Kubernetes监控方案之Grafana + Alertmanager + Prometheus_Kubernetes_12


3.1 Alertmanager简介 

Alertmanager 主要用于接收 Prometheus 发送的告警信息,它支持丰富的告警通知渠道,而且很容易做到告警信息进行去重,静默,分组等,将同类型的告警信息进行分类,防止告警信息轰炸,尽量精简告警的条数。 

3.2 Alertmanager容器化部署及应用

部署alertmanager-service.yaml:设置nodeport供外部访问。

[root@PH-K8S-M01 k8s-prometheus-grafana]# cat alertmanager-service.yaml

apiVersion: v1

kind: Service

metadata:

  name: alertmanager

  namespace: kube-system

  labels:

    kubernetes.io/cluster-service: "true"

    addonmanager.kubernetes.io/mode: Reconcile

    kubernetes.io/name: "Alertmanager"

spec:

  type: NodePort

  ports:

    - name: http

      port: 80

      protocol: TCP

      targetPort: 9093

      nodePort: 30004

  selector:

    k8s-app: alertmanager 

  #type: "ClusterIP"


部署alertmanager-templates.yaml:企业微信报警的短信模板,若使用默认模板,太过杂乱,不够精简,需要将该配置挂载至alertmanager-server中才会生效。告警模板中的变量都是通过调用prometheus pull到的metrics数据值获取到的

[root@PH-K8S-M01 k8s-prometheus-grafana]# cat alertmanager-templates.yaml 

apiVersion: v1

kind: ConfigMap

metadata:

    name: alertmanager-templates

    namespace: kube-system

data:

    wechat.tmpl: |

       {{ define "wechat.default.message" }}

       {{ if gt (len .Alerts.Firing) 0 -}}

       Alerts Firing:

       {{ range .Alerts}}

       告警级别: {{ .Labels.severity }}

       告警类型: {{ .Labels.alertname }}

       故障主机: {{ .Labels.instance }}

       告警主题: {{ .Annotations.summary }}

       告警详情: {{ .Annotations.description }}

       触发时间: {{ .StartsAt.Format "2006-01-02 15:04:05" }}

       {{- end }}

       {{- end }}

       {{ if gt (len .Alerts.Resolved) 0 -}}

       Alerts Resolved:

       {{ range .Alerts}}

       告警级别: {{ .Labels.severity }}

       告警类型: {{ .Labels.alertname }}

       故障主机: {{ .Labels.instance }}

       告警主题: {{ .Annotations.summary }}

       触发时间: {{ .StartsAt.Format "2006-01-02 15:04:05" }}

       恢复时间: {{ .EndsAt.Format "2006-01-02 15:04:05" }}

       {{- end }}

       {{- end }}

       {{- end }}


部署alertmanager-configmap.yaml:设置告警渠道,在这里配置了邮件告警,企业微信告警。在此设置告警时间间隔,分组,分流。

[root@PH-K8S-M01 k8s-prometheus-grafana]# cat alertmanager-configmap.yaml

apiVersion: v1

kind: ConfigMap

metadata:

  name: alertmanager-config

  namespace: kube-system

  labels:

    kubernetes.io/cluster-service: "true"

    addonmanager.kubernetes.io/mode: EnsureExists

data:

  alertmanager.yml: |

    global: 

      resolve_timeout: 5m    ##多久没有收到告警,默认告警已恢复

      smtp_smarthost: '*'    ##注意:这里的mail-server要加端口,要不然发不出去

      smtp_from: '*'

      smtp_auth_username: '*'

      smtp_auth_password: '*'


    templates:

    - '/etc/alertmanager-templates/*.tmpl'   ##指定告警模板路径


    receivers:

    - name: default-receiver      

      email_configs:

      - to: 'wujqc@*'

        send_resolved: true        ##是否发送告警恢复邮件

      wechat_configs:

      - send_resolved: true        ##是否发送告警恢复微信

        agent_id: '*'                ##应用页面上的配置

        to_party: '1'                ##企业微信后台->通讯录->部门ID

        corp_id: '*'                 ##应用页面上的配置

        api_secret: '*'              ##应用页面上的配置


    route:

      group_interval: 10s           ##第一次以后的告警分组要等到多久继续

      group_by: ['alertname']      ##根据prometheus-rule的alertname进行告警分组

      group_wait: 10s                ##第一次告警分组要等待多久

      receiver: default-receiver   ##告警组名称要与receivers保持一致

      repeat_interval: 5m           ##相同告警间隔多久发送下一条


agent_id,api_secret:

Kubernetes监控方案之Grafana + Alertmanager + Prometheus_Alertmanager_13


to_party:

Kubernetes监控方案之Grafana + Alertmanager + Prometheus_Kubernetes_14


corp_id

Kubernetes监控方案之Grafana + Alertmanager + Prometheus_Prometheus_15

Alertmanager告警可以根据标签值进行分流,如mysql告警分发到DBA组,架构告警发到架构组。实现方式如下,通过routes 的labels进行正则匹配,实现路由转发,若都不匹配则使用默认的route告警组default-receiver进行告警

route: 

receiver: 'default-receiver' 

group_wait: 30s 

group_interval: 5m 

repeat_interval: 4h 

group_by: [cluster, alertname] # All alerts that do not match the following child

routes

  # will remain at the root node and be dispatched to 'default-receiver'.

  routes:

  # All alerts with service=mysql or service=cassandra

  # are dispatched to the database pager.

  - receiver: 'database-pager'

    group_wait: 10s

    match_re:        ###正则匹配labels来判断使用哪个路由进行报警

      service: mysql|cassandra


部署alertmanager-deployment.yaml:部署alertmanager-server,挂载alertmanager configmap进行告警设置,在这里还在alertmanager-server pod创建的基础上加了配置文件热加载的镜像,保证重新加载alertmanager configmap配置时不需要重启alertmanager pod。

注意,这里还需要挂载alertmanager-templates,否则告警模板不生效

[root@PH-K8S-M01 k8s-prometheus-grafana]# cat alertmanager-deployment.yaml

apiVersion: apps/v1

kind: Deployment

metadata:

  name: alertmanager

  namespace: kube-system

  labels:

    k8s-app: alertmanager

    kubernetes.io/cluster-service: "true"

    addonmanager.kubernetes.io/mode: Reconcile

    version: v0.14.0

spec:

  replicas: 1

  selector:

    matchLabels:

      k8s-app: alertmanager

      version: v0.14.0

  template:

    metadata:

      labels:

        k8s-app: alertmanager

        version: v0.14.0

      annotations:

        scheduler.alpha.kubernetes.io/critical-pod: ''

    spec:

      priorityClassName: system-cluster-critical

      containers:

        - name: prometheus-alertmanager

          image: "prom/alertmanager:v0.14.0"

          imagePullPolicy: "IfNotPresent"

          args:

            - --config.file=/etc/config/alertmanager.yml

            - --storage.path=/data

            - --web.external-url=/

          ports:

            - containerPort: 9093

          readinessProbe:

            httpGet:

              path: /#/status

              port: 9093

            initialDelaySeconds: 30

            timeoutSeconds: 30

          volumeMounts:

            - name: config-volume

              mountPath: /etc/config

            - name: storage-volume

              mountPath: "/data"

              subPath: ""

            - name: templates-volume

              mountPath: /etc/alertmanager-templates   ##模板挂载点要与configmap一致

            - name: localtime

              mountPath: /etc/localtime


          resources:

            limits:

              cpu: 10m

              memory: 50Mi

            requests:

              cpu: 10m

              memory: 50Mi

        - name: prometheus-alertmanager-configmap-reload

          image: "jimmidyson/configmap-reload:v0.1"

          imagePullPolicy: "IfNotPresent"

          args:

            - --volume-dir=/etc/config

            - --webhook-url=http://localhost:9093/-/reload

          volumeMounts:

            - name: config-volume

              mountPath: /etc/config

              readOnly: true

          resources:

            limits:

              cpu: 10m

              memory: 10Mi

            requests:

              cpu: 10m

              memory: 10Mi

      volumes:

        - name: config-volume

          configMap:

            name: alertmanager-config

        - name: storage-volume

          emptyDir: {}

        - name: templates-volume

          configMap:

            name: alertmanager-templates     ###挂载告警模板

        - name: localtime

          hostPath:

            path: /etc/localtime


验证:可以手动触发prometheus-rule加载的配置,进行报警验证

邮件报警:

Fring表示触发告警

Resolved表示告警恢复

Kubernetes监控方案之Grafana + Alertmanager + Prometheus_Alertmanager_16


3.3 Alertmanager告警状态

Inactive:这里什么都没有发生。

Pending:已触发阈值,但未满足告警持续时间

Firing:已触发阈值且满足告警持续时间。警报发送给接受者。

Kubernetes监控方案之Grafana + Alertmanager + Prometheus_Kubernetes_17

3.4 Alertmanager告警静默设置

访问alertmanager后台,访问地址

http://IP:30004




Kubernetes监控方案之Grafana + Alertmanager + Prometheus_Alertmanager_18


右上角new silence用于手动设置静默,对一些需要忽略的告警设置维护模式,根据labels进行正则匹配设置,如


Kubernetes监控方案之Grafana + Alertmanager + Prometheus_Kubernetes_19