prometheus相关的服务都部署在monitoring这个namespace下
部署prometheus服务
-
创建namespace
$ cd /opt/k8s/prometheus $ cat>1-namespace.yml<<EOF apiVersion: v1 kind: Namespace metadata: name: monitoring EOF -
创建prometheus对应的配置文件，利用kubernetes的ConfigMap
$ cd /opt/k8s/prometheus $ cat>2-prom-cnfig.yml<<EOF apiVersion: v1 kind: ConfigMap metadata: name: prom-config namespace: monitoring data: prometheus.yml: | global: scrape_interval: 15s scrape_timeout: 15s scrape_configs: - job_name: 'prometheus' static_configs: - targets: ['localhost:9090'] EOF -
创建用来存储prometheus数据的pv、pvc（采用本地存储、可以通过搭建NFS、GlusterFs等分布式文件系统代替）
$ cd /opt/k8s/prometheus $ cat>3-prom-pv.yml<<EOF kind: PersistentVolume apiVersion: v1 metadata: namespace: monitoring name: prometheus labels: type: local app: prometheus spec: capacity: storage: 10Gi accessModes: - ReadWriteOnce hostPath: path: /opt/k8s/prometheus/data --- kind: PersistentVolumeClaim apiVersion: v1 metadata: namespace: monitoring name: prometheus-claim spec: accessModes: - ReadWriteOnce resources: requests: storage: 10Gi EOF -
创建启动prometheus的文件，以deployment形式部署，外部访问通过NodePort类型的service
$ cd /opt/k8s/prometheus $ cat>4-prometheus.yml<<EOF apiVersion: apps/v1 kind: Deployment metadata: name: prometheus namespace: monitoring labels: app: prometheus spec: selector: matchLabels: app: prometheus replicas: 1 template: metadata: labels: app: prometheus spec: containers: - name: prometheus image: prom/prometheus:v2.16.0 args: - '--config.file=/etc/prometheus/prometheus.yml' - '--storage.tsdb.path=/prometheus' - "--storage.tsdb.retention=7d" - "--web.enable-lifecycle" ports: - containerPort: 9090 volumeMounts: - mountPath: "/prometheus" subPath: prometheus name: data - mountPath: "/etc/prometheus" name: config resources: requests: cpu: 500m memory: 2Gi limits: cpu: 500m memory: 2Gi volumes: - name: config configMap: name: prom-config - name: data persistentVolumeClaim: claimName: prometheus-claim --- apiVersion: v1 kind: Service metadata: namespace: monitoring name: prometheus spec: type: NodePort ports: - port: 9090 targetPort: 9090 nodePort: 9090 selector: app: prometheus EOF- 在Prometheus的启动命令中，传入了参数storage.tsdb.path、storage.tsdb.retention分别指定了Prometheus数据存储路径、存储时间
- web.enable-lifecycle 当配置信息变更后通过/-/reload重新加载新的配置内容，不用重启服务
-
启动Prometheus服务
$ cd /opt/k8s/prometheus $ mkdir data & chmod -R 777 data $ kubectl create -f 1-namespace.yml -f 2-prom-cnfig.yml -f 3-prom-pv.yml -f 4-prometheus.yml -
查看组件状态，确保所有服务正常启动
$ kubectl get all -n monitoring NAME READY STATUS RESTARTS AGE pod/prometheus-57cf64764d-xqnvl 1/1 Running 0 51s NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE service/prometheus NodePort 10.254.209.164 <none> 9090:9090/TCP 51s NAME READY UP-TO-DATE AVAILABLE AGE deployment.apps/prometheus 1/1 1 1 51s NAME DESIRED CURRENT READY AGE replicaset.apps/prometheus-57cf64764d 1 1 1 51s -
界面访问
和kubernetes监控相关的技术
- cAdvisor google 开源的一款容器监控方案，收集容器自身的各种资源使用、性能相关信息，通过resutful API的形式提供给外部。kubernetes把cAdvisor对应的功能集成到了kubelet中，所以不需要再单独部署，直接从kubelet获取，并且kubelet还对容器信息进行summary，以pod为单位供外部调用。
- metrics-server 是kubernetes核心监控流程中的一个组件，是Heapster的替代方案，从kubelet指标接口中获取信息，主要是CPU、Memory，再通过API server暴露出去。主要供kubectl top、HPA等kubernetes组件使用。只在内存中存储最后一次获取到的指标信息，不负责数据存储
- kube-state-metrics 通过监听 API Server 生成有关资源对象的状态指标，比如 Deployment、replica sets等。只在内存中存储最后一次获取到的指标信息，不负责数据存储
- node-exporter Prometheus官方提供的，专门用来收集*NIX系统自身、以及对应硬件的指标信息
- kube-prometheus 一站式的kubernetes监控方案，将node-exporter、prometheus、kube-state-metrics、Grafana、metrics-server等组件收集起来，提供了更加便捷的脚本供使用者快速搭建一个完整的监控平台。
kubernetes监控内容
- 对集群自身状态的监控 ，如节点自身的CPU、Memory、IO、Network等信息
- kubernetes系统自身组件的监控如 kube-schedule-manager、kube-proxy、kubelet等
- 集群中运行容器的监控，如容器、Pod等为单元的CPU、Memory信息
- 集群中编排组件对应的指标监控，如Deployment、Daemonset等
本文采用自己部署组件的形式，看如何一步一步搭建一个监控平台。对于有快速搭建需求的，可以参考 kube-prometheus
部署node-exporter
因为要监控每一个节点，所以采用Daemonset控制器来部署node-exporter，在每个节点上都运行一个Pod
-
启动文件
$ cd /opt/k8s/prometheus $ cat>5-node-exporter.yml<<EOF apiVersion: apps/v1 kind: DaemonSet metadata: labels: app: node-exporter name: node-exporter namespace: monitoring spec: selector: matchLabels: app: node-exporter template: metadata: labels: app: node-exporter spec: containers: - name: node-exporter image: 192.168.0.107/prometheus/node-exporter:v0.18.1 args: - --web.listen-address=:9100 - --path.procfs=/host/proc - --path.sysfs=/host/sys - --path.sysfs=/host/sys - --path.rootfs=/host/root - --no-collector.hwmon - --collector.filesystem.ignored-mount-points=^/(dev|proc|sys|var/lib/docker/.+)($|/) - --collector.filesystem.ignored-fs-types=^(autofs|binfmt_misc|cgroup|configfs|debugfs|devpts|devtmpfs|fusectl|hugetlbfs|mqueue|overlay|proc|procfs|pstore|rpc_pipefs|securityfs|sysfs|tracefs)$ resources: limits: cpu: 250m memory: 180Mi requests: cpu: 102m memory: 180Mi ports: - containerPort: 9100 volumeMounts: - mountPath: /host/proc name: proc readOnly: false - mountPath: /host/sys name: sys readOnly: false - mountPath: /host/root mountPropagation: HostToContainer name: root readOnly: true hostNetwork: true hostPID: true nodeSelector: kubernetes.io/os: linux securityContext: runAsNonRoot: true runAsUser: 65534 tolerations: - operator: Exists volumes: - hostPath: path: /proc name: proc - hostPath: path: /sys name: sys - hostPath: path: / name: root EOF -
启动node-exporter
$ cd /opt/k8s/prometheus $ kubectl create -f 5-node-exporter.yml $ kubectl -n monitoring get pod | grep node node-exporter-854vr 1/1 Running 6 50m node-exporter-lv9pv 1/1 Running 0 50m -
通过prometheus收集node-exporter的指标信息
因为集群的节点之后可能会动态扩容和缩减，所以不便采用静态配置的形式，Prometheus给我们提供了Kubernetes对应的服务发现功能，可以实现对Kubernetes的动态监控。其中对节点的监控利用node的服务发现方式，在prometheus的配置文件中追加如下配置（对应的2-prom-cnfig.yml也需要追加，否则Configmap重建后这些信息就丢掉了）
$ kubectl -n monitoring edit configmaps prom-config- job_name: "kubernetes-nodes" kubernetes_sd_configs: - role: node relabel_configs: - source_labels: [__address__] regex: '(.*):10250' replacement: '${1}:9100' target_label: __address__ action: replace - action: labelmap regex: __meta_kubernetes_node_label_(.+)追加完成后执行如下命令，重新加载配置项，至于为何这样配置，后面章节配置原理具体讲解
$ curl -XPOST http://192.168.0.107:9090/-/reload此时，prometheus就会尝试获取集群node信息，查看promethes的日志信息，会发现如下错误提示
level=error ts=2020-03-22T10:37:13.856Z caller=klog.go:94 component=k8s_client_runtime func=ErrorDepth msg="/app/discovery/kubernetes/kubernetes.go:333: Failed to list *v1.Node: nodes is forbidden: User "system:serviceaccount:monitoring:default" cannot list resource "nodes" in API group "" at the cluster scope"意思是用默认的serviceaccount不能 list *v1.Node，因此需要我们为Prometheus重新创建serviceaccount，并赋予相应的权限
-
创建prometheus对应的serviceaccount,并赋予相应的权限
$ cd /opt/k8s/prometheus $ cat>6-prometheus-serivceaccount-role.yaml<<EOF apiVersion: v1 kind: ServiceAccount metadata: name: prometheus-k8s namespace: monitoring --- apiVersion: rbac.authorization.k8s.io/v1 kind: ClusterRole metadata: name: prometheus-k8s rules: - apiGroups: [""] resources: - nodes/proxy - nodes - namespaces - endpoints - pods - services verbs: ["get","list","watch"] - apiGroups: [""] resources: - nodes/metrics verbs: ["get"] - nonResourceURLs: - /metrics verbs: ["get"] - apiGroups: - extensions resources: - ingresses verbs: ["get", "list", "watch"] --- apiVersion: rbac.authorization.k8s.io/v1 kind: ClusterRoleBinding metadata: name: prometheus-k8s roleRef: apiGroup: rbac.authorization.k8s.io kind: ClusterRole name: prometheus-k8s subjects: - kind: ServiceAccount name: prometheus-k8s namespace: monitoring EOF$ cd /opt/k8s/prometheus $ kubectl create -f 6-prometheus-serivceaccount-role.yaml修改Prometheus的启动yaml，追加serivceaccount配置,重启Prometheus
... spec: serviceAccountName: prometheus-k8s containers: - name: prometheus ...查看prometheus监控对象列表
kubernetes_sd_config 配置原理详解
-
目标地址查找
当kubernetes_sd_config配置的role是node时，prometheus启动后会调用kubernetes的 LIST Node API获取node相关信息，并从node对象中获取IP和PORT来构成监控地址。
其中IP获取按照如下顺序InternalIP, ExternalIP, LegacyHostIP, HostName查找
Port值默认采用Kubelet的HTTP port。
通过如下命令可查看 list Node接口返回的IP 和 Port信息
$ kubectl get node -o=jsonpath='{range .items[*]}{.status.addresses}{" "}{.status.daemonEndpoints}{" "}{end}' [map[address:192.168.0.107 type:InternalIP] map[address:master type:Hostname]] map[kubeletEndpoint:map[Port:10250]] [map[address:192.168.0.114 type:InternalIP] map[address:slave type:Hostname]] map[kubeletEndpoint:map[Port:10250]]上述返回结果表示集群中有两个节点，构成的taget地址分别是
192.168.0.107:10250 192.168.0.114:10250 -
relabe_configs
Relabeling可以让Prometheus在抓取数据之前动态的修改标签的值。Prometheus有许多默认标签，其中下面几个和我们处理有关
__address__:初始化时会设置成目标地址对应的<host>:<port>instance:__address__标签的值经过Relabel阶段后，会设置给标签instance, 即instance是__address__标签经过Relabel后的值__scheme__:默认值 http__metrics_path__默认值 /metrics
Prometheus拉取指标信息的目的地址是把这几个标签连接起来
__scheme__://instance/__metrics_path__ -
我们启动node-expeorter后，在各个节点的
:9100/metrics上暴露了node指标信息，然后在prometheus中追加了如下配置段- job_name: "kubernetes-nodes" kubernetes_sd_configs: - role: node relabel_configs: - source_labels: [__address__] regex: '(.*):10250' replacement: '${1}:9100' target_label: __address__ action: replace - action: labelmap regex: __meta_kubernetes_node_label_(.+)其中relabel_configs的第一个配置片段
- source_labels: [__address__] regex: '(.*):10250' replacement: '${1}:9100' target_label: __address__ action: replace- 通过regex从
__address__中匹配出IP地址 - replacement：对应的值设置成
${IP}:9100 - target_label: 将
__address__的值替换成replacement，即${IP}:9100
经过这些步骤后，拼接成的获取指标地址为
[http://192.168.0.107:9100/metrics, http://192.168.0.114:9100/metrics],和我们的node-exporter暴露的指标地址匹配，就可以拉取Node的指标信息了另外因为prometheus会将node对应的标签变成
__meta_kubernetes_node_label_<labelname>，所以追加了一个labelmap的动作，再把这些标签名字还原出来 - 通过regex从
-
完整配置例子参考prometheus-kubernetes
追加收集kubelete提供的指标
kubelet会收集一些api server 、etcd等服务的指标信息,可以通过如下命令查看
$ kubectl get --raw https://192.168.0.107:10250/metrics
- 其中10250是kubelet的默认监听端口
接下来通过在promehteus中追加配置，让promehteus拉取这些信息
- job_name: "kubernetes-kubelet"
scheme: https
tls_config:
ca_file: /var/run/secrets/kubernetes.io/serviceaccount/ca.crt
bearer_token_file: /var/run/secrets/kubernetes.io/serviceaccount/token
kubernetes_sd_configs:
- role: node
relabel_configs:
- action: labelmap
regex: __meta_kubernetes_node_label_(.+)
追加收集cAdvisor指标,实现对集群容器的监控
kubelet默认集成了cAdvisor，用来对集群中容器信息进行收集，Kubernetes 1.7.3以及之后的版本中，把cAdvisor收集的指标信息(以container_开头）从Kubelet对应的/metrics中移除了，所以需要额外配置一个收集cAdvisor的job。cAdvisor指标调用命令
$ kubectl get --raw https://192.168.0.107:6443/api/v1/nodes/master/proxy/metrics/cadvisor
- 其中10250是kubelet的默认监听端口
接下来通过在promehteus中追加配置，让promehteus拉取这些信息
- job_name: "kubernetes-cadvisor"
scheme: https
tls_config:
ca_file: /var/run/secrets/kubernetes.io/serviceaccount/ca.crt
bearer_token_file: /var/run/secrets/kubernetes.io/serviceaccount/token
kubernetes_sd_configs:
- role: node
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
添加完成后重新load 配置文件，查看prometheus监控对象列表
完整的配置文件如下
$ cd /opt/k8s/prometheus
$ cat 2-prom-cnfig.yml
apiVersion: v1
kind: ConfigMap
metadata:
name: prom-config
namespace: monitoring
data:
prometheus.yml: |
global:
scrape_interval: 15s
scrape_timeout: 15s
scrape_configs:
- job_name: 'prometheus'
static_configs:
- targets: ['localhost:9090']
- job_name: "kubernetes-nodes"
kubernetes_sd_configs:
- role: node
relabel_configs:
- source_labels: [__address__]
regex: '(.*):10250'
replacement: '${1}:9100'
target_label: __address__
action: replace
- action: labelmap
regex: __meta_kubernetes_node_label_(.+)
- job_name: "kubernetes-kubelet"
scheme: https
tls_config:
ca_file: /var/run/secrets/kubernetes.io/serviceaccount/ca.crt
bearer_token_file: /var/run/secrets/kubernetes.io/serviceaccount/token
kubernetes_sd_configs:
- role: node
relabel_configs:
- action: labelmap
regex: __meta_kubernetes_node_label_(.+)
- job_name: "kubernetes-cadvisor"
scheme: https
tls_config:
ca_file: /var/run/secrets/kubernetes.io/serviceaccount/ca.crt
bearer_token_file: /var/run/secrets/kubernetes.io/serviceaccount/token
kubernetes_sd_configs:
- role: node
relabel_configs:
- action: labelmap
regex: __meta_kubernetes_node_label_(.+)
- target_label: __address__
replacement: kubernetes.default.svc:443
- source_labels: [__meta_kubernetes_node_name]
regex: (.+)
target_label: __metrics_path__
replacement: /api/v1/nodes/${1}/proxy/metrics/cadvisor
部署grafana
-
创建用来存储grafana数据的pv、pvc（采用本地存储、可以通过搭建NFS、GlusterFs等分布式文件系统代替）
$ cd /opt/k8s/prometheus $ cat>7-grafana-pv.yml<<EOF kind: PersistentVolume apiVersion: v1 metadata: namespace: monitoring name: grafana labels: type: local app: grafana spec: capacity: storage: 10Gi accessModes: - ReadWriteOnce hostPath: path: /opt/k8s/prometheus/grafana-pvc --- kind: PersistentVolumeClaim apiVersion: v1 metadata: namespace: monitoring name: grafana-claim spec: accessModes: - ReadWriteOnce resources: requests: storage: 10Gi EOF -
grafana部署文件
$ cd /opt/k8s/prometheus $ cat>8-grafana.yml<<EOF apiVersion: apps/v1 kind: Deployment metadata: labels: app: grafana name: grafana namespace: monitoring spec: replicas: 1 selector: matchLabels: app: grafana template: metadata: labels: app: grafana spec: containers: - image: grafana/grafana:6.6.2 name: grafana ports: - containerPort: 3000 name: http readinessProbe: httpGet: path: /api/health port: http resources: limits: cpu: 200m memory: 400Mi requests: cpu: 100m memory: 200Mi volumeMounts: - mountPath: /var/lib/grafana name: grafana-pvc readOnly: false subPath: data - mountPath: /etc/grafana/provisioning/datasources name: grafana-pvc readOnly: false subPath: datasources - mountPath: /etc/grafana/provisioning/dashboards name: grafana-pvc readOnly: false subPath: dashboards-pro - mountPath: /grafana-dashboard-definitions/0 name: grafana-pvc readOnly: false subPath: dashboards nodeSelector: beta.kubernetes.io/os: linux securityContext: runAsNonRoot: true runAsUser: 65534 volumes: - name: grafana-pvc persistentVolumeClaim: claimName: grafana-claim --- apiVersion: v1 kind: Service metadata: namespace: monitoring name: grafana spec: type: NodePort ports: - port: 3000 targetPort: 3000 nodePort: 3000 selector: app: grafana EOF -
启动grafana
-
创建grafana挂载目录
$ cd /opt/k8s/prometheus $ mkdir -p grafana-pvc/data $ mkdir -p grafana-pvc/datasources $ mkdir -p grafana-pvc/dashboards-pro $ mkdir -p grafana-pvc/dashboards $ chmod -R 777 grafana-pvc- data目录存放grafana的数据
- datasources存放预定义的数据源
- dashboards-pro存放dashboards管理文件，其中配置的dashboard的文件地址指向grafana-pvc/dashboards挂载到容器中的地址/grafana-dashboard-definitions/0
- dashboards存放真正的dashboards定义文件（json）
-
创建默认的数据源文件
$ cd /opt/k8s/prometheus/grafana-pvc/datasources $ cat > datasource.yaml<<EOF apiVersion: 1 datasources: - name: Prometheus type: prometheus access: proxy url: http://prometheus.monitoring.svc:9090 EOF -
创建默认的dashboards管理文件
$ cd /opt/k8s/prometheus/grafana-pvc/dashboards-pro $ cat >dashboards.yaml<<EOF apiVersion: 1 providers: - name: '0' orgId: 1 folder: '' type: file editable: true updateIntervalSeconds: 10 allowUiUpdates: false options: path: /grafana-dashboard-definitions/0 EOF -
创建默认的dashboard定义文件
可以到 a collection of shared dashboards,找到自己需要的dashboard模版，下载对应的json文件将对应文件存放到/opt/k8s/prometheus/grafana-pvc/dashboards）,这里作为示例，下载1 Node Exporter for Prometheus Dashboard CN v20191102,对应的ID是8919。
$ cd /opt/k8s/prometheus/grafana-pvc/dashboards $ wget https://grafana.com/api/dashboards/8919/revisions/11/download -o node-exporter-k8s.json因为模版中的数据源默认用的是
${DS_PROMETHEUS_111},从界面导入时有配置项供替换，我们直接下载json文件，所以通过直接修改文件，把数据源改成我们在/opt/k8s/prometheus/grafana-pvc/datasources下配置的数据源$ cd /opt/k8s/prometheus/grafana-pvc/dashboards $ sed -i "s/${DS_PROMETHEUS_111}/Prometheus/g" node-exporter-k8s.json修改title
... "timezone": "browser", "title": "k8s-node-monitoring", ... -
启动
$ cd /opt/k8s/prometheus/ $ kubectl create -f 7-grafana-pv.yml 8-grafana.yml
-
-
通过界面查看，因为我们已经默认设置过数据源、dashboard等信息，可以直接查看对应的dashboard
在部署grafana时，我们配置了默认的数据源、dashboard等信息，主要是为了实现，系统部署后这些默认监控指标可以直接观察，不需要实施人员现场配置。
其他的监控例如使用 Kube-state-metrics以及cAdvisor metrics实现对集群中Deployment、StatefulSet、容器、pod的监控也可以采用这种形式来实现。如可以利用1. Kubernetes Deployment Statefulset Daemonset metrics作为模版，稍微修改满足我们的监控需要，这里就不再展示具体步骤，读者可以自行尝试。