目录

一、自动发现类型
1、Node
2、Service
3、Pod
4、endpoints
5、ingress
二、自动发现Pod配置
1、创建发现规则
2、创建对应的Secret对象
3、promethues添加资源对象
4、应用配置
5、添加权限

一、自动发现类型

在上一篇文中留了一个坑

 

监控某个statefulset服务的时候,我在service文件中定义了个EP,然后把pod的ip写死在配置文件中,这样,当pod重启后,IP地址变化,就监控不到数据了,这肯定是不合理的。

如果在我们的 Kubernetes 集群中有了很多的 Service/Pod,那么我们都需要一个一个的去建立一个对应的 ServiceMonitor 对象来进行监控吗?这样岂不是也很麻烦么?

为解决上面的问题,Prometheus Operator 为我们提供了一个额外的抓取配置的来解决这个问题,我们可以通过额外的配置来获取k8s的资源监控(pod、service、node等)。

promethues支持多种文件发现。

 

其中通过kubernetes_sd_configs,可以达到我们想要的目的,监控其各种资源。kubernetes SD 配置允许从kubernetes REST API接受搜集指标,且总是和集群保持同步状态,以下任何一种role类型都能够配置来发现我们想要的对象,来自官网翻译的。

1、Node

Node role发现每个集群中的目标是通过默认的kubelet的HTTP端口。目标地址默认是kubernetes如下地址中node的第一个地址(NodeInternalIPNodeExternalIP,NodeLegacyHostIP, and NodeHostName.)

可用的meta标签有:

1
2
3
4
5
6
__meta_kubernetes_node_name: The name of the node object.
__meta_kubernetes_node_label_<labelname>: Each label from the node object.
__meta_kubernetes_node_labelpresent_<labelname>: true for each label from the node object.
__meta_kubernetes_node_annotation_<annotationname>: Each annotation from the node object.
__meta_kubernetes_node_annotationpresent_<annotationname>: true for each annotation from the node object.
__meta_kubernetes_node_address_<address_type>: The first address for each node address type, if it exists.

此外,node的实例标签将会被设置成从API server传递过来的node的name。

2、Service

service角色会为每个服务发现一个服务端口。对于黑盒监控的服务,这个比较有用。address将会被设置成service的kubernetes DNS名称以及各自的服务端口。

Available meta labels:

1
2
3
4
5
6
7
8
9
10
__meta_kubernetes_namespace: The namespace of the service object.
__meta_kubernetes_service_annotation_<annotationname>: Each annotation from the service object.
__meta_kubernetes_service_annotationpresent_<annotationname>: "true" for each annotation of the service object.
__meta_kubernetes_service_cluster_ip: The cluster IP address of the service. (Does not apply to services of type ExternalName)
__meta_kubernetes_service_external_name: The DNS name of the service. (Applies to services of type ExternalName)
__meta_kubernetes_service_label_<labelname>: Each label from the service object.
__meta_kubernetes_service_labelpresent_<labelname>: true for each label of the service object.
__meta_kubernetes_service_name: The name of the service object.
__meta_kubernetes_service_port_name: Name of the service port for the target.
__meta_kubernetes_service_port_protocol: Protocol of the service port for the target.

 

3、Pod

Pod role会发现所有pods以及暴露的容器作为target。每个容器声明一个端口,一个单独的target就会生成。如果一个容器没有指定端口,通过relabel手动指定一个端口,一个port-free target容器将会生成。

Available meta labels:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
__meta_kubernetes_namespace: The namespace of the pod object.
__meta_kubernetes_pod_name: The name of the pod object.
__meta_kubernetes_pod_ip: The pod IP of the pod object.
__meta_kubernetes_pod_label_<labelname>: Each label from the pod object.
__meta_kubernetes_pod_labelpresent_<labelname>: truefor each label from the pod object.
__meta_kubernetes_pod_annotation_<annotationname>: Each annotation from the pod object.
__meta_kubernetes_pod_annotationpresent_<annotationname>: true for each annotation from the pod object.
__meta_kubernetes_pod_container_init: true if the container is an InitContainer
__meta_kubernetes_pod_container_name: Name of the container the target address points to.
__meta_kubernetes_pod_container_port_name: Name of the container port.
__meta_kubernetes_pod_container_port_number: Number of the container port.
__meta_kubernetes_pod_container_port_protocol: Protocol of the container port.
__meta_kubernetes_pod_ready: Set to true or false for the pod's ready state.
__meta_kubernetes_pod_phase: Set to Pending, Running, Succeeded, Failed or Unknown in the lifecycle.
__meta_kubernetes_pod_node_name: The name of the node the pod is scheduled onto.
__meta_kubernetes_pod_host_ip: The current host IP of the pod object.
__meta_kubernetes_pod_uid: The UID of the pod object.
__meta_kubernetes_pod_controller_kind: Object kind of the pod controller.
__meta_kubernetes_pod_controller_name: Name of the pod controller.

 

4、endpoints

endpoints role从每个服务监听的endpoints发现。每个endpoint都会发现一个port。如果endpoint是一个pod,所有包含的容器不被绑定到一个endpoint port,也会被targets被发现。

Available meta labels:

1
2
3
4
5
6
7
8
9
10
11
12
__meta_kubernetes_namespace: The namespace of the endpoints object.
__meta_kubernetes_endpoints_name: The names of the endpoints object.
For all targets discovered directly from the endpoints list (those not additionally inferred from underlying pods), the following labels are attached:
__meta_kubernetes_endpoint_hostname: Hostname of the endpoint.
__meta_kubernetes_endpoint_node_name: Name of the node hosting the endpoint.
__meta_kubernetes_endpoint_ready: Set to true or false for the endpoint's ready state.
__meta_kubernetes_endpoint_port_name: Name of the endpoint port.
__meta_kubernetes_endpoint_port_protocol: Protocol of the endpoint port.
__meta_kubernetes_endpoint_address_target_kind: Kind of the endpoint address target.
__meta_kubernetes_endpoint_address_target_name: Name of the endpoint address target.
If the endpoints belong to a service, all labels of the role: service discovery are attached.
For all targets backed by a pod, all labels of the role: pod discovery are attached.

 

5、ingress

ingress role将会发现每个ingress。ingress在黑盒监控上比较有用。address将会被设置成ingress指定的配置。

Available meta labels:

1
2
3
4
5
6
7
8
__meta_kubernetes_namespace: The namespace of the ingress object.
__meta_kubernetes_ingress_name: The name of the ingress object.
__meta_kubernetes_ingress_label_<labelname>: Each label from the ingress object.
__meta_kubernetes_ingress_labelpresent_<labelname>: true for each label from the ingress object.
__meta_kubernetes_ingress_annotation_<annotationname>: Each annotation from the ingress object.
__meta_kubernetes_ingress_annotationpresent_<annotationname>: true for each annotation from the ingress object.
__meta_kubernetes_ingress_scheme: Protocol scheme of ingress, https if TLS config is set. Defaults to http.
__meta_kubernetes_ingress_path: Path from ingress spec. Defaults to /.

 

二、自动发现Pod配置

比如业务上有一个微服务,类型为statefulset,启动后是2个pod的副本集,pod暴露的数据接口为http://pod_ip:7000/metrics。由于pod每次重启后,ip都会变化,所以只能通过自动发现的方式获取数据。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
apiVersion: apps/v1
kind: StatefulSet
metadata:
  labels:
    run: jx3recipe
  name: jx3recipe
  annotations:
    prometheus.io/scrape: "true"
spec:
  selector:
    matchLabels:
      app: jx3recipe
  serviceName: jx3recipe-service
  replicas: 2
  template:
    metadata:
      labels:
        app: jx3recipe
        appCluster: jx3recipe-cluster
    spec:
      terminationGracePeriodSeconds: 20
      containers:
      - image: hub.kce.ooo.com/jx3pvp/jx3recipe:qa-latest
        imagePullPolicy: Always
        securityContext:
          runAsUser: 1000
        name: jx3recipe
        lifecycle:
          preStop:
            exec:
              command: ["kill","-s","SIGINT","1"]
        volumeMounts:
        - name: config-volume
          mountPath: /data/conf.yml
          subPath: conf.yml
        resources:
          requests:
            cpu: "100m"
            memory: "500Mi"
        env:
        - name: JX3PVP_ENV
          value: "qa"
        - name: JX3PVP_RUN_MODE
          value: "k8s"
        - name: JX3PVP_SERVICE_ID
          valueFrom:
            fieldRef:
              fieldPath: metadata.name
        - name: JX3PVP_LOCAL_IP
          valueFrom:
            fieldRef:
              fieldPath: status.podIP
        - name: JX3PVP_CONSUL_IP
          value: $(CONSUL_AGENT_SERVICE_HOST)
        ports:
            - name: biz
              containerPort: 8000
              protocol: "TCP"
            - name: admin
              containerPort: 7000
              protocol: "TCP"
      volumes:
        - name: config-volume
          configMap:
            name: app-configure-file-jx3recipe
            items:
            - key: jx3recipe.yml
              path: conf.yml

 

1、创建发现规则

设定发现pod规则:文件名为promethues-additional.yaml

  • pod名称的label为jx3recipe
  • pod的label_appCluster匹配为 jx3recipe-cluster
  • pod的address为http://.*:7000/metrics格式
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
- job_name: 'kubernetes-service-pod'
  kubernetes_sd_configs:
  - role: pod
  relabel_configs:
  - source_labels: [__meta_kubernetes_pod_container_name]
    action: replace
    target_label: jx3recipe
  - action: labelmap
    regex: __meta_kubernetes_pod_label_(.+)
  - source_labels:  ["__meta_kubernetes_pod_label_appCluster"]
    regex: "jx3recipe-cluster"
    action: keep
  - source_labels: [__address__]
    action: keep
    regex: '(.*):7000'

 

2、创建对应的Secret对象

1
kubectl create secret generic additional-configs --from-file=prometheus-additional.yaml -n monitoring

 

创建完成后,会将上面配置信息进行 base64 编码后作为 prometheus-additional.yaml 这个 key 对应的值存在:

1
2
3
4
5
6
7
8
9
10
11
12
apiVersion: v1
data:
  prometheus-additional.yaml: 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
kind: Secret
metadata:
  creationTimestamp: "2019-09-10T09:32:22Z"
  name: additional-configs
  namespace: monitoring
  resourceVersion: "1004681"
  selfLink: /api/v1/namespaces/monitoring/secrets/additional-configs
  uid: e455d657-d3ad-11e9-95b4-fa163e3c10ff
type: Opaque

然后我们只需要在声明 prometheus 的资源对象文件中添加上这个额外的配置:(prometheus-prometheus.yaml)

3、promethues添加资源对象

修改prometheus-prometheus.yaml文件

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
apiVersion: monitoring.coreos.com/v1
kind: Prometheus
metadata:
  labels:
    prometheus: k8s
  name: k8s
  namespace: monitoring
spec:
  alerting:
    alertmanagers:
    - name: alertmanager-main
      namespace: monitoring
      port: web
  baseImage: quay.io/prometheus/prometheus
  nodeSelector:
    beta.kubernetes.io/os: linux
  replicas: 2
  secrets:
  - etcd-certs
  resources:
    requests:
      memory: 400Mi
  ruleSelector:
    matchLabels:
      prometheus: k8s
      role: alert-rules
  securityContext:
    fsGroup: 2000
    runAsNonRoot: true
    runAsUser: 1000
  additionalScrapeConfigs:
    name: additional-configs
    key: prometheus-additional.yaml
  serviceAccountName: prometheus-k8s
  serviceMonitorNamespaceSelector: {}
  serviceMonitorSelector: {}
  version: v2.5.0

增加了下面这一段:

1
2
3
additionalScrapeConfigs:
    name: additional-configs
    key: prometheus-additional.yaml

 

4、应用配置

1
kubectl apply -f prometheus-prometheus.yaml

过一段时间,刷新promethues上的config,将会看到下面红色框框的配置。

 

 

5、添加权限

在 Prometheus Dashboard 的配置页面下面我们可以看到已经有了对应的的配置信息了,但是我们切换到 targets 页面下面却并没有发现对应的监控任务,查看 Prometheus 的 Pod 日志:

 

可以看到有很多错误日志出现,都是xxx is forbidden,这说明是 RBAC 权限的问题,通过 prometheus 资源对象的配置可以知道 Prometheus 绑定了一个名为 prometheus-k8s 的 ServiceAccount 对象,而这个对象绑定的是一个名为 prometheus-k8s 的 ClusterRole:(prometheus-clusterRole.yaml)

修改为:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  name: prometheus-k8s
rules:
- apiGroups:
  ""
  resources:
  - nodes
  - services
  - endpoints
  - pods
  - nodes/proxy
  verbs:
  get
  - list
  - watch
- apiGroups:
  ""
  resources:
  - configmaps
  - nodes/metrics
  verbs:
  get
- nonResourceURLs:
  - /metrics
  verbs:
  get

更新上面的 ClusterRole 这个资源对象,然后重建下 Prometheus 的所有 Pod,正常就可以看到 targets 页面下面有 kubernetes-service-pod这个监控任务了:

至此,一个自动发现pod的配置就完成了,其他资源(service、endpoint、ingress、node同样也可以通过自动发现的方式实现。)

发表评论

邮箱地址不会被公开。 必填项已用*标注