kubernetes常见日志采集问题和解决方案分析

传统日志与kubernetes日志对比

• 传统服务

1. 目录固定
2. 重启不会丢失
3. 不用关注标准与错误日志输出

• 容器服务

1. 节点不固定
2. 重启服务会漂移
3. 需要关注标准与错误日志输出
4. 日志文件重启会丢失
5. 日志目录不固定

k8s中的日志处理

常见日志采集方案：

1.远程日志 将容器中日志直接写入远程kafka、es等，再由logstash等处理。劣势就是需要改造服务由写入本地的就要进行修改 传输到远端存储。

2.sidecar模式，在每个pod中运行一个filebeat，logstash等pod共享一个valume，由采集工具将日志内容采集发送。缺点每个pod都需要额外增加一个日志采集工具，对pod有侵入。

3.logagent模式，在node中运行一个filebeat，logstash等，通过将pod日志目录固定事先约定好，直接采集对应目录内容传输至远端。优点节约资源，对pod无侵入。缺点：文件后缀名尽量统一否则维护较为困难。目录预先定义好无法判断来源于哪个pod，只能区分node。

实践：

采用logagent模式

日志采集工具使用阿里开源logpilot（优点与docker主进程进行交互能够通过docker动态识别各个pod日志目录，底层采用filebeat）

常见日志采集工具 logpilot filebeat logstash  Fluentd Logtail Flume（filebeat与flentd在容器环境中应用最多）缺点：都是采集监测静态目录。

搭建es：

---
apiVersion: v1
kind: Service
metadata:
  name: elasticsearch-api
  namespace: kube-system
  labels:
    name: elasticsearch
spec:
  selector:
    app: es
  ports:
  - name: transport
    port: 9200
    protocol: TCP
---
apiVersion: v1
kind: Service
metadata:
  name: elasticsearch-discovery
  namespace: kube-system
  labels:
    name: elasticsearch
spec:
  selector:
    app: es
  ports:
  - name: transport
    port: 9300
    protocol: TCP
---
apiVersion: apps/v1beta1
kind: StatefulSet
metadata:
  name: elasticsearch
  namespace: kube-system
  labels:
    kubernetes.io/cluster-service: "true"
spec:
  replicas: 3
  serviceName: "elasticsearch-service"
  selector:
    matchLabels:
      app: es
  template:
    metadata:
      labels:
        app: es
    spec:
      tolerations:
      - effect: NoSchedule
        key: node-role.kubernetes.io/master
      serviceAccountName: dashboard-admin
      initContainers:
      - name: init-sysctl
        image: busybox:1.27
        command:
        - sysctl
        - -w
        - vm.max_map_count=262144
        securityContext:
          privileged: true
      containers:
      - name: elasticsearch
        image: registry.cn-hangzhou.aliyuncs.com/imooc/elasticsearch:5.5.1
        ports:
        - containerPort: 9200
          protocol: TCP
        - containerPort: 9300
          protocol: TCP
        securityContext:
          capabilities:
            add:
              - IPC_LOCK
              - SYS_RESOURCE
        resources:
          limits:
            memory: 4000Mi
          requests:
            cpu: 100m
            memory: 2000Mi
        env:
          - name: "http.host"
            value: "0.0.0.0"
          - name: "network.host"
            value: "_eth0_"
          - name: "cluster.name"
            value: "docker-cluster"
          - name: "bootstrap.memory_lock"
            value: "false"
          - name: "discovery.zen.ping.unicast.hosts"
            value: "elasticsearch-discovery"
          - name: "discovery.zen.ping.unicast.hosts.resolve_timeout"
            value: "10s"
          - name: "discovery.zen.ping_timeout"
            value: "6s"
          - name: "discovery.zen.minimum_master_nodes"
            value: "2"
          - name: "discovery.zen.fd.ping_interval"
            value: "2s"
          - name: "discovery.zen.no_master_block"
            value: "write"
          - name: "gateway.expected_nodes"
            value: "2"
          - name: "gateway.expected_master_nodes"
            value: "1"
          - name: "transport.tcp.connect_timeout"
            value: "60s"
          - name: "ES_JAVA_OPTS"
            value: "-Xms2g -Xmx2g"
        livenessProbe:
          tcpSocket:
            port: transport
          initialDelaySeconds: 20
          periodSeconds: 10
        volumeMounts:
        - name: es-data
          mountPath: /data
      terminationGracePeriodSeconds: 30
      volumes:
      - name: es-data
        hostPath:
          path: /es-data

es.yaml

搭建logpilot：

---
apiVersion: extensions/v1beta1
kind: DaemonSet
metadata:
  name: log-pilot
  namespace: kube-system
  labels:
    k8s-app: log-pilot
    kubernetes.io/cluster-service: "true"
spec:
  template:
    metadata:
      labels:
        k8s-app: log-es
        kubernetes.io/cluster-service: "true"
        version: v1.22
    spec:
      tolerations:
      - key: node-role.kubernetes.io/master
        effect: NoSchedule
      serviceAccountName: dashboard-admin
      containers:
      - name: log-pilot
        image: registry.cn-hangzhou.aliyuncs.com/imooc/log-pilot:0.9-filebeat
        resources:
          limits:
            memory: 200Mi
          requests:
            cpu: 100m
            memory: 200Mi
        env:
          - name: "FILEBEAT_OUTPUT"
            value: "elasticsearch"
          - name: "ELASTICSEARCH_HOST"
            value: "elasticsearch-api"
          - name: "ELASTICSEARCH_PORT"
            value: "9200"
          - name: "ELASTICSEARCH_USER"
            value: "elastic"
          - name: "ELASTICSEARCH_PASSWORD"
            value: "changeme"
        volumeMounts:
        - name: sock
          mountPath: /var/run/docker.sock
        - name: root
          mountPath: /host
          readOnly: true
        - name: varlib
          mountPath: /var/lib/filebeat
        - name: varlog
          mountPath: /var/log/filebeat
        securityContext:
          capabilities:
            add:
            - SYS_ADMIN
      terminationGracePeriodSeconds: 30
      volumes:
      - name: sock
        hostPath:
          path: /var/run/docker.sock
      - name: root
        hostPath:
          path: /
      - name: varlib
        hostPath:
          path: /var/lib/filebeat
          type: DirectoryOrCreate
      - name: varlog
        hostPath:
          path: /var/log/filebeat
          type: DirectoryOrCreate

logpilot.yaml

搭建kibana：

---
apiVersion: v1
kind: Service
metadata:
  name: kibana
  namespace: kube-system
  labels:
    component: kibana
spec:
  selector:
    component: kibana
  ports:
  - name: http
    port: 80
    targetPort: http
---
#ingress
apiVersion: extensions/v1beta1
kind: Ingress
metadata:
  name: kibana
  namespace: kube-system
spec:
  rules:
  - host: kibana.mooc.com
    http:
      paths:
      - path: /
        backend:
          serviceName: kibana
          servicePort: 80
---
apiVersion: apps/v1beta1
kind: Deployment
metadata:
  name: kibana
  namespace: kube-system
  labels:
    component: kibana
spec:
  replicas: 1
  selector:
    matchLabels:
     component: kibana
  template:
    metadata:
      labels:
        component: kibana
    spec:
      containers:
      - name: kibana
        image: registry.cn-hangzhou.aliyuncs.com/imooc/kibana:5.5.1
        env:
        - name: CLUSTER_NAME
          value: docker-cluster
        - name: ELASTICSEARCH_URL
          value: http://elasticsearch-api:9200/
        resources:
          limits:
            cpu: 1000m
          requests:
            cpu: 100m
        ports:
        - containerPort: 5601
          name: http

kibana.yaml