Flink on k8s 讲解与实战操作

一、概述

Flink核心是一个流式的数据流执行引擎,并且能够基于同一个Flink运行时,提供支持流处理和批处理两种类型应用。其针对数据流的分布式计算提供了数据分布,数据通信及容错机制等功能。

  • Flink官网:https://flink.apache.org/

  • 不同版本的文档:https://nightlies.apache.org/flink/

  • k8s on flink 官方文档:https://nightlies.apache.org/flink/flink-docs-release-1.14/docs/deployment/resource-providers/native_kubernetes/

  • GitHub地址:https://github.com/apache/flink/tree/release-1.14.6/

二、Flink 运行模式

官方文档:https://nightlies.apache.org/flink/flink-docs-release-1.15/zh/docs/deployment/overview/

FLink on yarn 有三种运行模式:

  • yarn-session模式(Seesion Mode)

  • yarn-cluster模式(Per-Job Mode)

  • Application模式(Application Mode)

Flink on k8s 讲解与实战操作_第1张图片

【温馨提示】Per-Job 模式(已弃用),Per-job 模式仅由 YARN 支持,并已在 Flink 1.15 中弃用。它将被丢弃在FLINK-26000中。

三、Flink on k8s实战操作

Flink on k8s 讲解与实战操作_第2张图片

1)flink下载

下载地址:https://flink.apache.org/downloads.html

wget https://dlcdn.apache.org/flink/flink-1.14.6/flink-1.14.6-bin-scala_2.12.tgz

2)构建基础镜像

docker pull apache/flink:1.14.6-scala_2.12
docker tag apache/flink:1.14.6-scala_2.12 myharbor.com/bigdata/flink:1.14.6-scala_2.12
docker push myharbor.com/bigdata/flink:1.14.6-scala_2.12

3)session模式

Flink Session 集群作为长时间运行的 Kubernetes Deployment 执行。你可以在一个Session 集群上运行多个 Flink 作业。每个作业都需要在集群部署完成后提交到集群。 Kubernetes 中的Flink Session 集群部署至少包含三个组件:

  • 运行JobManager的部署

  • TaskManagers池的部署

  • 暴露JobManager 的REST 和 UI 端口的服务

1、Native Kubernetes 模式

参数配置: https://nightlies.apache.org/flink/flink-docs-release-1.14/docs/deployment/config/#kubernetes-namespace

【1】构建镜像Dockerfile

FROM myharbor.com/bigdata/flink:1.14.6-scala_2.12
RUN rm -f /etc/localtime && ln -sv /usr/share/zoneinfo/Asia/Shanghai /etc/localtime && echo "Asia/Shanghai" > /etc/timezone
RUN export LANG=zh_CN.UTF-8

开始构建镜像

docker build -t myharbor.com/bigdata/flink-session:1.14.6-scala_2.12 . --no-cache

# 上传镜像
docker push myharbor.com/bigdata/flink-session:1.14.6-scala_2.12

【2】创建命名空间和serviceaccount

# 创建namespace
kubectl create ns flink
# 创建serviceaccount
kubectl create serviceaccount flink-service-account -n flink
# 用户授权
kubectl create clusterrolebinding flink-role-binding-flink --clusterrole=edit --serviceaccount=flink:flink-service-account

【3】创建flink集群

./bin/kubernetes-session.sh \
 -Dkubernetes.cluster-id=my-first-flink-cluster  \
 -Dkubernetes.container.image=myharbor.com/bigdata/flink-session:1.14.6-scala_2.12 \
 -Dkubernetes.namespace=flink \
 -Dkubernetes.jobmanager.service-account=flink-service-account \
 -Dkubernetes.rest-service.exposed.type=NodePort

Flink on k8s 讲解与实战操作_第3张图片

【4】提交任务

./bin/flink run \
    --target kubernetes-session \
    -Dkubernetes.cluster-id=my-first-flink-cluster \
    -Dkubernetes.namespace=flink \
    -Dkubernetes.jobmanager.service-account=flink-service-account \
    ./examples/streaming/TopSpeedWindowing.jar
    
    
    #   参数配置
    ./examples/streaming/WordCount.jar
    -Dkubernetes.taskmanager.cpu=2000m \
    -Dexternal-resource.limits.kubernetes.cpu=4000m \
 -Dexternal-resource.limits.kubernetes.memory=10Gi \
 -Dexternal-resource.requests.kubernetes.cpu=2000m \
 -Dexternal-resource.requests.kubernetes.memory=8Gi \
 -Dkubernetes.taskmanager.cpu=2000m \

【温馨提示】注意jdk版本,目前jdk8是正常的。

Flink on k8s 讲解与实战操作_第4张图片

【5】查看

kubectl get pods -n flink
kubectl logs -f my-first-flink-cluster-taskmanager-1-1

Flink on k8s 讲解与实战操作_第5张图片

Flink on k8s 讲解与实战操作_第6张图片

【6】删除flink集群

kubectl delete deployment/my-first-flink-cluster -n flink
kubectl delete ns flink --force

2、Standalone模式

【1】构建镜像

默认用户是flink用户,这里我换成admin,根据企业需要更换用户,脚本可以通过上面运行的pod拿到。

启动脚本 docker-entrypoint.sh

#!/usr/bin/env bash

###############################################################################
#  Licensed to the Apache Software Foundation (ASF) under one
#  or more contributor license agreements.  See the NOTICE file
#  distributed with this work for additional information
#  regarding copyright ownership.  The ASF licenses this file
#  to you under the Apache License, Version 2.0 (the
#  "License"); you may not use this file except in compliance
#  with the License.  You may obtain a copy of the License at
#
#      http://www.apache.org/licenses/LICENSE-2.0
#
#  Unless required by applicable law or agreed to in writing, software
#  distributed under the License is distributed on an "AS IS" BASIS,
#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
#  See the License for the specific language governing permissions and
# limitations under the License.
###############################################################################

COMMAND_STANDALONE="standalone-job"
COMMAND_HISTORY_SERVER="history-server"

# If unspecified, the hostname of the container is taken as the JobManager address
JOB_MANAGER_RPC_ADDRESS=${JOB_MANAGER_RPC_ADDRESS:-$(hostname -f)}
CONF_FILE="${FLINK_HOME}/conf/flink-conf.yaml"

drop_privs_cmd() {
    if [ $(id -u) != 0 ]; then
        # Don't need to drop privs if EUID != 0
        return
    elif [ -x /sbin/su-exec ]; then
        # Alpine
        echo su-exec admin
    else
        # Others
        echo gosu admin
    fi
}

copy_plugins_if_required() {
  if [ -z "$ENABLE_BUILT_IN_PLUGINS" ]; then
    return 0
  fi

  echo "Enabling required built-in plugins"
  for target_plugin in $(echo "$ENABLE_BUILT_IN_PLUGINS" | tr ';' ' '); do
    echo "Linking ${target_plugin} to plugin directory"
    plugin_name=${target_plugin%.jar}

    mkdir -p "${FLINK_HOME}/plugins/${plugin_name}"
    if [ ! -e "${FLINK_HOME}/opt/${target_plugin}" ]; then
      echo "Plugin ${target_plugin} does not exist. Exiting."
      exit 1
    else
      ln -fs "${FLINK_HOME}/opt/${target_plugin}" "${FLINK_HOME}/plugins/${plugin_name}"
      echo "Successfully enabled ${target_plugin}"
    fi
  done
}

set_config_option() {
  local option=$1
  local value=$2

  # escape periods for usage in regular expressions
  local escaped_option=$(echo ${option} | sed -e "s/\./\\\./g")

  # either override an existing entry, or append a new one
  if grep -E "^${escaped_option}:.*" "${CONF_FILE}" > /dev/null; then
        sed -i -e "s/${escaped_option}:.*/$option: $value/g" "${CONF_FILE}"
  else
        echo "${option}: ${value}" >> "${CONF_FILE}"
  fi
}

prepare_configuration() {
    set_config_option jobmanager.rpc.address ${JOB_MANAGER_RPC_ADDRESS}
    set_config_option blob.server.port 6124
    set_config_option query.server.port 6125

    if [ -n "${TASK_MANAGER_NUMBER_OF_TASK_SLOTS}" ]; then
        set_config_option taskmanager.numberOfTaskSlots ${TASK_MANAGER_NUMBER_OF_TASK_SLOTS}
    fi

    if [ -n "${FLINK_PROPERTIES}" ]; then
        echo "${FLINK_PROPERTIES}" >> "${CONF_FILE}"
    fi
    envsubst < "${CONF_FILE}" > "${CONF_FILE}.tmp" && mv "${CONF_FILE}.tmp" "${CONF_FILE}"
}

maybe_enable_jemalloc() {
    if [ "${DISABLE_JEMALLOC:-false}" == "false" ]; then
        JEMALLOC_PATH="/usr/lib/$(uname -m)-linux-gnu/libjemalloc.so"
        JEMALLOC_FALLBACK="/usr/lib/x86_64-linux-gnu/libjemalloc.so"
        if [ -f "$JEMALLOC_PATH" ]; then
            export LD_PRELOAD=$LD_PRELOAD:$JEMALLOC_PATH
        elif [ -f "$JEMALLOC_FALLBACK" ]; then
            export LD_PRELOAD=$LD_PRELOAD:$JEMALLOC_FALLBACK
        else
            if [ "$JEMALLOC_PATH" = "$JEMALLOC_FALLBACK" ]; then
                MSG_PATH=$JEMALLOC_PATH
            else
                MSG_PATH="$JEMALLOC_PATH and $JEMALLOC_FALLBACK"
            fi
            echo "WARNING: attempted to load jemalloc from $MSG_PATH but the library couldn't be found. glibc will be used instead."
        fi
    fi
}

maybe_enable_jemalloc

copy_plugins_if_required

prepare_configuration

args=("$@")
if [ "$1" = "help" ]; then
    printf "Usage: $(basename "$0") (jobmanager|${COMMAND_STANDALONE}|taskmanager|${COMMAND_HISTORY_SERVER})\n"
    printf "    Or $(basename "$0") help\n\n"
    printf "By default, Flink image adopts jemalloc as default memory allocator. This behavior can be disabled by setting the 'DISABLE_JEMALLOC' environment variable to 'true'.\n"
    exit 0
elif [ "$1" = "jobmanager" ]; then
    args=("${args[@]:1}")

    echo "Starting Job Manager"

    exec $(drop_privs_cmd) "$FLINK_HOME/bin/jobmanager.sh" start-foreground "${args[@]}"
elif [ "$1" = ${COMMAND_STANDALONE} ]; then
    args=("${args[@]:1}")

    echo "Starting Job Manager"

    exec $(drop_privs_cmd) "$FLINK_HOME/bin/standalone-job.sh" start-foreground "${args[@]}"
elif [ "$1" = ${COMMAND_HISTORY_SERVER} ]; then
    args=("${args[@]:1}")

    echo "Starting History Server"

    exec $(drop_privs_cmd) "$FLINK_HOME/bin/historyserver.sh" start-foreground "${args[@]}"
elif [ "$1" = "taskmanager" ]; then
    args=("${args[@]:1}")

    echo "Starting Task Manager"

    exec $(drop_privs_cmd) "$FLINK_HOME/bin/taskmanager.sh" start-foreground "${args[@]}"
fi

args=("${args[@]}")

# Running command in pass-through mode
exec $(drop_privs_cmd) "${args[@]}"

编排Dockerfile

FROM myharbor.com/bigdata/centos:7.9.2009

USER root

# 安装常用工具
RUN yum install -y vim tar wget curl rsync bzip2 iptables tcpdump less telnet net-tools lsof

# 设置时区,默认是UTC时区
RUN rm -f /etc/localtime && ln -sv /usr/share/zoneinfo/Asia/Shanghai /etc/localtime && echo "Asia/Shanghai" > /etc/timezone

RUN mkdir -p /opt/apache

ADD jdk-8u212-linux-x64.tar.gz /opt/apache/

ADD flink-1.14.6-bin-scala_2.12.tgz  /opt/apache/

ENV FLINK_HOME /opt/apache/flink-1.14.6
ENV JAVA_HOME /opt/apache/jdk1.8.0_212
ENV PATH $JAVA_HOME/bin:$PATH

# 创建用户应用jar目录
RUN mkdir $FLINK_HOME/usrlib/

#RUN mkdir home
COPY docker-entrypoint.sh /opt/apache/
RUN chmod +x /opt/apache/docker-entrypoint.sh

RUN groupadd --system --gid=9999 admin && useradd --system --home-dir $FLINK_HOME --uid=9999 --gid=admin admin

RUN chown -R admin:admin /opt/apache

#设置的工作目录
WORKDIR $FLINK_HOME

# 对外暴露端口
EXPOSE 6123 8081

# 执行脚本,构建镜像时不执行,运行实例才会执行
ENTRYPOINT ["/opt/apache/docker-entrypoint.sh"]
CMD ["help"]

开始构建镜像

docker build -t myharbor.com/bigdata/flink-centos-admin:1.14.6-scala_2.12 . --no-cache

# 上传镜像
docker push myharbor.com/bigdata/flink-centos-admin:1.14.6-scala_2.12

# 删除镜像
docker rmi myharbor.com/bigdata/flink-centos-admin:1.14.6-scala_2.12
crictl rmi myharbor.com/bigdata/flink-centos-admin:1.14.6-scala_2.12

【2】创建命名空间和serviceaccount

# 创建namespace
kubectl create ns flink
# 创建serviceaccount
kubectl create serviceaccount flink-service-account -n flink
# 用户授权
kubectl create clusterrolebinding flink-role-binding-flink --clusterrole=edit --serviceaccount=flink:flink-service-account

【3】编排yaml文件

  • flink-configuration-configmap.yaml

apiVersion: v1
kind: ConfigMap
metadata:
  name: flink-config
  labels:
    app: flink
data:
  flink-conf.yaml: |+
    jobmanager.rpc.address: flink-jobmanager
    taskmanager.numberOfTaskSlots: 2
    blob.server.port: 6124
    jobmanager.rpc.port: 6123
    taskmanager.rpc.port: 6122
    queryable-state.proxy.ports: 6125
    jobmanager.memory.process.size: 3200m
    taskmanager.memory.process.size: 2728m
    taskmanager.memory.flink.size: 2280m
    parallelism.default: 2    
  log4j-console.properties: |+
    # This affects logging for both user code and Flink
    rootLogger.level = INFO
    rootLogger.appenderRef.console.ref = ConsoleAppender
    rootLogger.appenderRef.rolling.ref = RollingFileAppender

    # Uncomment this if you want to _only_ change Flink's logging
    #logger.flink.name = org.apache.flink
    #logger.flink.level = INFO

    # The following lines keep the log level of common libraries/connectors on
    # log level INFO. The root logger does not override this. You have to manually
    # change the log levels here.
    logger.akka.name = akka
    logger.akka.level = INFO
    logger.kafka.name= org.apache.kafka
    logger.kafka.level = INFO
    logger.hadoop.name = org.apache.hadoop
    logger.hadoop.level = INFO
    logger.zookeeper.name = org.apache.zookeeper
    logger.zookeeper.level = INFO

    # Log all infos to the console
    appender.console.name = ConsoleAppender
    appender.console.type = CONSOLE
    appender.console.layout.type = PatternLayout
    appender.console.layout.pattern = %d{yyyy-MM-dd HH:mm:ss,SSS} %-5p %-60c %x - %m%n

    # Log all infos in the given rolling file
    appender.rolling.name = RollingFileAppender
    appender.rolling.type = RollingFile
    appender.rolling.append = false
    appender.rolling.fileName = ${sys:log.file}
    appender.rolling.filePattern = ${sys:log.file}.%i
    appender.rolling.layout.type = PatternLayout
    appender.rolling.layout.pattern = %d{yyyy-MM-dd HH:mm:ss,SSS} %-5p %-60c %x - %m%n
    appender.rolling.policies.type = Policies
    appender.rolling.policies.size.type = SizeBasedTriggeringPolicy
    appender.rolling.policies.size.size=100MB
    appender.rolling.strategy.type = DefaultRolloverStrategy
    appender.rolling.strategy.max = 10

    # Suppress the irrelevant (wrong) warnings from the Netty channel handler
    logger.netty.name = org.jboss.netty.channel.DefaultChannelPipeline
    logger.netty.level = OFF
  • jobmanager-service.yaml可选服务,仅非 HA 模式需要。

apiVersion: v1
kind: Service
metadata:
  name: flink-jobmanager
spec:
  type: ClusterIP
  ports:
  - name: rpc
    port: 6123
  - name: blob-server
    port: 6124
  - name: webui
    port: 8081
  selector:
    app: flink
    component: jobmanager
  • jobmanager-rest-service.yaml 可选服务,将 jobmanager rest端口公开为公共 Kubernetes 节点的端口。

apiVersion: v1
kind: Service
metadata:
  name: flink-jobmanager-rest
spec:
  type: NodePort
  ports:
  - name: rest
    port: 8081
    targetPort: 8081
    nodePort: 30081
  selector:
    app: flink
    component: jobmanager
  • taskmanager-query-state-service.yaml 可选服务,公开 TaskManager 端口以访问可查询状态作为公共 Kubernetes 节点的端口。

apiVersion: v1
kind: Service
metadata:
  name: flink-taskmanager-query-state
spec:
  type: NodePort
  ports:
  - name: query-state
    port: 6125
    targetPort: 6125
    nodePort: 30025
  selector:
    app: flink
    component: taskmanager

以上几个配置文件是公共的

  • jobmanager-session-deployment-non-ha.yaml

apiVersion: apps/v1
kind: Deployment
metadata:
  name: flink-jobmanager
spec:
  replicas: 1
  selector:
    matchLabels:
      app: flink
      component: jobmanager
  template:
    metadata:
      labels:
        app: flink
        component: jobmanager
    spec:
      containers:
      - name: jobmanager
        image: myharbor.com/bigdata/flink-centos-admin:1.14.6-scala_2.12
        args: ["jobmanager"]
        ports:
        - containerPort: 6123
          name: rpc
        - containerPort: 6124
          name: blob-server
        - containerPort: 8081
          name: webui
        livenessProbe:
          tcpSocket:
            port: 6123
          initialDelaySeconds: 30
          periodSeconds: 60
        volumeMounts:
        - name: flink-config-volume
          mountPath: /opt/apache/flink-1.14.6/conf/
        securityContext:
          runAsUser: 9999  # refers to user _flink_ from official flink image, change if necessary
      volumes:
      - name: flink-config-volume
        configMap:
          name: flink-config
          items:
          - key: flink-conf.yaml
            path: flink-conf.yaml
          - key: log4j-console.properties
            path: log4j-console.properties
  • taskmanager-session-deployment.yaml

apiVersion: apps/v1
kind: Deployment
metadata:
  name: flink-taskmanager
spec:
  replicas: 2
  selector:
    matchLabels:
      app: flink
      component: taskmanager
  template:
    metadata:
      labels:
        app: flink
        component: taskmanager
    spec:
      containers:
      - name: taskmanager
        image: myharbor.com/bigdata/flink-centos-admin:1.14.6-scala_2.12
        args: ["taskmanager"]
        ports:
        - containerPort: 6122
          name: rpc
        - containerPort: 6125
          name: query-state
        livenessProbe:
          tcpSocket:
            port: 6122
          initialDelaySeconds: 30
          periodSeconds: 60
        volumeMounts:
        - name: flink-config-volume
          mountPath: /opt/apache/flink-1.14.6/conf/
        securityContext:
          runAsUser: 9999  # refers to user _flink_ from official flink image, change if necessary
      volumes:
      - name: flink-config-volume
        configMap:
          name: flink-config
          items:
          - key: flink-conf.yaml
            path: flink-conf.yaml
          - key: log4j-console.properties
            path: log4j-console.properties

【4】创建flink集群

kubectl create ns flink
# Configuration and service definition
kubectl create -f flink-configuration-configmap.yaml -n flink

# service
kubectl create -f jobmanager-service.yaml -n flink
kubectl create -f jobmanager-rest-service.yaml -n flink
kubectl create -f taskmanager-query-state-service.yaml -n flink

# Create the deployments for the cluster
kubectl create -f jobmanager-session-deployment-non-ha.yaml -n flink
kubectl create -f taskmanager-session-deployment.yaml -n flink

镜像逆向解析dockerfile

alias whaler="docker run -t --rm -v /var/run/docker.sock:/var/run/docker.sock:ro pegleg/whaler"
whaler flink:1.14.6-scala_2.12

查看

kubectl get pods,svc -n flink -owide

Flink on k8s 讲解与实战操作_第7张图片

web:http://192.168.182.110:30081/#/overview

Flink on k8s 讲解与实战操作_第8张图片

【5】提交任务

./bin/flink run -m local-168-182-110:30081 ./examples/streaming/WordCount.jar

Flink on k8s 讲解与实战操作_第9张图片

kubectl logs flink-taskmanager-54649bf96c-zjtkh -n flink

Flink on k8s 讲解与实战操作_第10张图片

Flink on k8s 讲解与实战操作_第11张图片

【6】删除flink集群

kubectl delete -f jobmanager-service.yaml -n flink
kubectl delete -f flink-configuration-configmap.yaml -n flink
kubectl delete -f taskmanager-session-deployment.yaml -n flink
kubectl delete -f jobmanager-session-deployment.yaml -n flink
kubectl delete ns flink --force

【7】访问flink web

端口就是jobmanager-rest-service.yaml文件中的NodePort

http://192.168.182.110:30081/#/overview

Flink on k8s 讲解与实战操作_第12张图片

4)application模式(推荐)

Kubernetes 中一个基本的Flink Application 集群部署包含三个组件:

  • 运行JobManager的应用程序

  • TaskManagers池的部署

  • 暴露JobManager 的REST 和 UI 端口的服务

1、Native Kubernetes 模式(常用)

【1】构建镜像Dockerfile

FROM myharbor.com/bigdata/flink:1.14.6-scala_2.12
RUN rm -f /etc/localtime && ln -sv /usr/share/zoneinfo/Asia/Shanghai /etc/localtime && echo "Asia/Shanghai" > /etc/timezone
RUN export LANG=zh_CN.UTF-8
RUN mkdir -p $FLINK_HOME/usrlib
COPY  TopSpeedWindowing.jar $FLINK_HOME/usrlib/

开始构建镜像

docker build -t myharbor.com/bigdata/flink-application:1.14.6-scala_2.12 . --no-cache

# 上传镜像
docker push myharbor.com/bigdata/flink-application:1.14.6-scala_2.12

# 删除镜像
docker rmi myharbor.com/bigdata/flink-application:1.14.6-scala_2.12
crictl rmi myharbor.com/bigdata/flink-application:1.14.6-scala_2.12

【2】创建命名空间和serviceacount

# 创建namespace
kubectl create ns flink
# 创建serviceaccount
kubectl create serviceaccount flink-service-account -n flink
# 用户授权
kubectl create clusterrolebinding flink-role-binding-flink --clusterrole=edit --serviceaccount=flink:flink-service-account

【3】创建flink集群并提交任务

./bin/flink run-application \
    --target kubernetes-application \
    -Dkubernetes.cluster-id=my-first-application-cluster  \
 -Dkubernetes.container.image=myharbor.com/bigdata/flink-application:1.14.6-scala_2.12 \
 -Dkubernetes.jobmanager.replicas=1 \
 -Dkubernetes.namespace=flink \
 -Dkubernetes.jobmanager.service-account=flink-service-account \
 -Dexternal-resource.limits.kubernetes.cpu=2000m \
 -Dexternal-resource.limits.kubernetes.memory=2Gi \
 -Dexternal-resource.requests.kubernetes.cpu=1000m \
 -Dexternal-resource.requests.kubernetes.memory=1Gi \
 -Dkubernetes.rest-service.exposed.type=NodePort \
 local:///opt/flink/usrlib/TopSpeedWindowing.jar

【注意】 local是应用模式中唯一支持的方案。local代表本地环境,这里即pod或者容器环境,并非宿主机。

查看

kubectl get pods pods,svc -n flink

Flink on k8s 讲解与实战操作_第13张图片

kubectl logs -f my-first-application-cluster-taskmanager-1-1 -n flink

Flink on k8s 讲解与实战操作_第14张图片

Flink on k8s 讲解与实战操作_第15张图片

Flink on k8s 讲解与实战操作_第16张图片

【4】删除flink集群

kubectl delete deployment/my-first-application-cluster -n flink
kubectl delete ns flink --force

2、Standalone模式

【1】构建镜像 Dockerfile

启动脚本 docker-entrypoint.sh

#!/usr/bin/env bash

###############################################################################
#  Licensed to the Apache Software Foundation (ASF) under one
#  or more contributor license agreements.  See the NOTICE file
#  distributed with this work for additional information
#  regarding copyright ownership.  The ASF licenses this file
#  to you under the Apache License, Version 2.0 (the
#  "License"); you may not use this file except in compliance
#  with the License.  You may obtain a copy of the License at
#
#      http://www.apache.org/licenses/LICENSE-2.0
#
#  Unless required by applicable law or agreed to in writing, software
#  distributed under the License is distributed on an "AS IS" BASIS,
#  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
#  See the License for the specific language governing permissions and
# limitations under the License.
###############################################################################

COMMAND_STANDALONE="standalone-job"
COMMAND_HISTORY_SERVER="history-server"

# If unspecified, the hostname of the container is taken as the JobManager address
JOB_MANAGER_RPC_ADDRESS=${JOB_MANAGER_RPC_ADDRESS:-$(hostname -f)}
CONF_FILE="${FLINK_HOME}/conf/flink-conf.yaml"

drop_privs_cmd() {
    if [ $(id -u) != 0 ]; then
        # Don't need to drop privs if EUID != 0
        return
    elif [ -x /sbin/su-exec ]; then
        # Alpine
        echo su-exec admin
    else
        # Others
        echo gosu admin
    fi
}

copy_plugins_if_required() {
  if [ -z "$ENABLE_BUILT_IN_PLUGINS" ]; then
    return 0
  fi

  echo "Enabling required built-in plugins"
  for target_plugin in $(echo "$ENABLE_BUILT_IN_PLUGINS" | tr ';' ' '); do
    echo "Linking ${target_plugin} to plugin directory"
    plugin_name=${target_plugin%.jar}

    mkdir -p "${FLINK_HOME}/plugins/${plugin_name}"
    if [ ! -e "${FLINK_HOME}/opt/${target_plugin}" ]; then
      echo "Plugin ${target_plugin} does not exist. Exiting."
      exit 1
    else
      ln -fs "${FLINK_HOME}/opt/${target_plugin}" "${FLINK_HOME}/plugins/${plugin_name}"
      echo "Successfully enabled ${target_plugin}"
    fi
  done
}

set_config_option() {
  local option=$1
  local value=$2

  # escape periods for usage in regular expressions
  local escaped_option=$(echo ${option} | sed -e "s/\./\\\./g")

  # either override an existing entry, or append a new one
  if grep -E "^${escaped_option}:.*" "${CONF_FILE}" > /dev/null; then
        sed -i -e "s/${escaped_option}:.*/$option: $value/g" "${CONF_FILE}"
  else
        echo "${option}: ${value}" >> "${CONF_FILE}"
  fi
}

prepare_configuration() {
    set_config_option jobmanager.rpc.address ${JOB_MANAGER_RPC_ADDRESS}
    set_config_option blob.server.port 6124
    set_config_option query.server.port 6125

    if [ -n "${TASK_MANAGER_NUMBER_OF_TASK_SLOTS}" ]; then
        set_config_option taskmanager.numberOfTaskSlots ${TASK_MANAGER_NUMBER_OF_TASK_SLOTS}
    fi

    if [ -n "${FLINK_PROPERTIES}" ]; then
        echo "${FLINK_PROPERTIES}" >> "${CONF_FILE}"
    fi
    envsubst < "${CONF_FILE}" > "${CONF_FILE}.tmp" && mv "${CONF_FILE}.tmp" "${CONF_FILE}"
}

maybe_enable_jemalloc() {
    if [ "${DISABLE_JEMALLOC:-false}" == "false" ]; then
        JEMALLOC_PATH="/usr/lib/$(uname -m)-linux-gnu/libjemalloc.so"
        JEMALLOC_FALLBACK="/usr/lib/x86_64-linux-gnu/libjemalloc.so"
        if [ -f "$JEMALLOC_PATH" ]; then
            export LD_PRELOAD=$LD_PRELOAD:$JEMALLOC_PATH
        elif [ -f "$JEMALLOC_FALLBACK" ]; then
            export LD_PRELOAD=$LD_PRELOAD:$JEMALLOC_FALLBACK
        else
            if [ "$JEMALLOC_PATH" = "$JEMALLOC_FALLBACK" ]; then
                MSG_PATH=$JEMALLOC_PATH
            else
                MSG_PATH="$JEMALLOC_PATH and $JEMALLOC_FALLBACK"
            fi
            echo "WARNING: attempted to load jemalloc from $MSG_PATH but the library couldn't be found. glibc will be used instead."
        fi
    fi
}

maybe_enable_jemalloc

copy_plugins_if_required

prepare_configuration

args=("$@")
if [ "$1" = "help" ]; then
    printf "Usage: $(basename "$0") (jobmanager|${COMMAND_STANDALONE}|taskmanager|${COMMAND_HISTORY_SERVER})\n"
    printf "    Or $(basename "$0") help\n\n"
    printf "By default, Flink image adopts jemalloc as default memory allocator. This behavior can be disabled by setting the 'DISABLE_JEMALLOC' environment variable to 'true'.\n"
    exit 0
elif [ "$1" = "jobmanager" ]; then
    args=("${args[@]:1}")

    echo "Starting Job Manager"

    exec $(drop_privs_cmd) "$FLINK_HOME/bin/jobmanager.sh" start-foreground "${args[@]}"
elif [ "$1" = ${COMMAND_STANDALONE} ]; then
    args=("${args[@]:1}")

    echo "Starting Job Manager"

    exec $(drop_privs_cmd) "$FLINK_HOME/bin/standalone-job.sh" start-foreground "${args[@]}"
elif [ "$1" = ${COMMAND_HISTORY_SERVER} ]; then
    args=("${args[@]:1}")

    echo "Starting History Server"

    exec $(drop_privs_cmd) "$FLINK_HOME/bin/historyserver.sh" start-foreground "${args[@]}"
elif [ "$1" = "taskmanager" ]; then
    args=("${args[@]:1}")

    echo "Starting Task Manager"

    exec $(drop_privs_cmd) "$FLINK_HOME/bin/taskmanager.sh" start-foreground "${args[@]}"
fi

args=("${args[@]}")

# Running command in pass-through mode
exec $(drop_privs_cmd) "${args[@]}"

编排Dockerfile

FROM myharbor.com/bigdata/centos:7.9.2009

USER root

# 安装常用工具
RUN yum install -y vim tar wget curl rsync bzip2 iptables tcpdump less telnet net-tools lsof

# 设置时区,默认是UTC时区
RUN rm -f /etc/localtime && ln -sv /usr/share/zoneinfo/Asia/Shanghai /etc/localtime && echo "Asia/Shanghai" > /etc/timezone

RUN mkdir -p /opt/apache

ADD jdk-8u212-linux-x64.tar.gz /opt/apache/

ADD flink-1.14.6-bin-scala_2.12.tgz  /opt/apache/

ENV FLINK_HOME /opt/apache/flink-1.14.6
ENV JAVA_HOME /opt/apache/jdk1.8.0_212
ENV PATH $JAVA_HOME/bin:$PATH

# 创建用户应用jar目录
RUN mkdir $FLINK_HOME/usrlib/

#RUN mkdir home
COPY docker-entrypoint.sh /opt/apache/

RUN groupadd --system --gid=9999 admin && useradd --system --home-dir $FLINK_HOME --uid=9999 --gid=admin admin

RUN chown -R admin:admin /opt/apache
RUN chmod +x ${FLINK_HOME}/docker-entrypoint.sh

#设置的工作目录
WORKDIR $FLINK_HOME

# 对外暴露端口
EXPOSE 6123 8081

# 执行脚本,构建镜像时不执行,运行实例才会执行
ENTRYPOINT ["/opt/apache/docker-entrypoint.sh"]
CMD ["help"]
docker build -t myharbor.com/bigdata/flink-centos-admin:1.14.6-scala_2.12 . --no-cache

# 上传镜像
docker push myharbor.com/bigdata/flink-centos-admin:1.14.6-scala_2.12

# 删除镜像
docker rmi myharbor.com/bigdata/flink-centos-admin:1.14.6-scala_2.12

【2】创建命名空间和 serviceacount

# 创建namespace
kubectl create ns flink
# 创建serviceaccount
kubectl create serviceaccount flink-service-account -n flink
# 用户授权
kubectl create clusterrolebinding flink-role-binding-flink --clusterrole=edit --serviceaccount=flink:flink-service-account

【3】编排yaml文件

flink-configuration-configmap.yaml
apiVersion: v1
kind: ConfigMap
metadata:
  name: flink-config
  labels:
    app: flink
data:
  flink-conf.yaml: |+
    jobmanager.rpc.address: flink-jobmanager
    taskmanager.numberOfTaskSlots: 2
    blob.server.port: 6124
    jobmanager.rpc.port: 6123
    taskmanager.rpc.port: 6122
    queryable-state.proxy.ports: 6125
    jobmanager.memory.process.size: 3200m
    taskmanager.memory.process.size: 2728m
    taskmanager.memory.flink.size: 2280m
    parallelism.default: 2    
  log4j-console.properties: |+
    # This affects logging for both user code and Flink
    rootLogger.level = INFO
    rootLogger.appenderRef.console.ref = ConsoleAppender
    rootLogger.appenderRef.rolling.ref = RollingFileAppender

    # Uncomment this if you want to _only_ change Flink's logging
    #logger.flink.name = org.apache.flink
    #logger.flink.level = INFO

    # The following lines keep the log level of common libraries/connectors on
    # log level INFO. The root logger does not override this. You have to manually
    # change the log levels here.
    logger.akka.name = akka
    logger.akka.level = INFO
    logger.kafka.name= org.apache.kafka
    logger.kafka.level = INFO
    logger.hadoop.name = org.apache.hadoop
    logger.hadoop.level = INFO
    logger.zookeeper.name = org.apache.zookeeper
    logger.zookeeper.level = INFO

    # Log all infos to the console
    appender.console.name = ConsoleAppender
    appender.console.type = CONSOLE
    appender.console.layout.type = PatternLayout
    appender.console.layout.pattern = %d{yyyy-MM-dd HH:mm:ss,SSS} %-5p %-60c %x - %m%n

    # Log all infos in the given rolling file
    appender.rolling.name = RollingFileAppender
    appender.rolling.type = RollingFile
    appender.rolling.append = false
    appender.rolling.fileName = ${sys:log.file}
    appender.rolling.filePattern = ${sys:log.file}.%i
    appender.rolling.layout.type = PatternLayout
    appender.rolling.layout.pattern = %d{yyyy-MM-dd HH:mm:ss,SSS} %-5p %-60c %x - %m%n
    appender.rolling.policies.type = Policies
    appender.rolling.policies.size.type = SizeBasedTriggeringPolicy
    appender.rolling.policies.size.size=100MB
    appender.rolling.strategy.type = DefaultRolloverStrategy
    appender.rolling.strategy.max = 10

    # Suppress the irrelevant (wrong) warnings from the Netty channel handler
    logger.netty.name = org.jboss.netty.channel.DefaultChannelPipeline
    logger.netty.level = OFF

jobmanager-service.yaml可选服务,仅非 HA 模式需要。

apiVersion: v1
kind: Service
metadata:
  name: flink-jobmanager
spec:
  type: ClusterIP
  ports:
  - name: rpc
    port: 6123
  - name: blob-server
    port: 6124
  - name: webui
    port: 8081
  selector:
    app: flink
    component: jobmanager

jobmanager-rest-service.yaml 可选服务,将 jobmanager rest端口公开为公共 Kubernetes 节点的端口。

apiVersion: v1
kind: Service
metadata:
  name: flink-jobmanager-rest
spec:
  type: NodePort
  ports:
  - name: rest
    port: 8081
    targetPort: 8081
    nodePort: 30081
  selector:
    app: flink
    component: jobmanager

taskmanager-query-state-service.yaml 可选服务,公开 TaskManager 端口以访问可查询状态作为公共 Kubernetes 节点的端口。

apiVersion: v1
kind: Service
metadata:
  name: flink-taskmanager-query-state
spec:
  type: NodePort
  ports:
  - name: query-state
    port: 6125
    targetPort: 6125
    nodePort: 30025
  selector:
    app: flink
    component: taskmanager

jobmanager-application-non-ha.yaml ,非高可用

apiVersion: batch/v1
kind: Job
metadata:
  name: flink-jobmanager
spec:
  template:
    metadata:
      labels:
        app: flink
        component: jobmanager
    spec:
      restartPolicy: OnFailure
      containers:
        - name: jobmanager
          image: myharbor.com/bigdata/flink-centos-admin:1.14.6-scala_2.12
          env:
          args: ["standalone-job", "--job-classname", "org.apache.flink.examples.java.wordcount.WordCount","--output","/tmp/result"]
          ports:
            - containerPort: 6123
              name: rpc
            - containerPort: 6124
              name: blob-server
            - containerPort: 8081
              name: webui
          livenessProbe:
            tcpSocket:
              port: 6123
            initialDelaySeconds: 30
            periodSeconds: 60
          volumeMounts:
            - name: flink-config-volume
              mountPath: /opt/apache/flink-1.14.6/conf
            - name: job-artifacts-volume
              mountPath: /opt/apache/flink-1.14.6/usrlib
          securityContext:
            runAsUser: 9999  # refers to user _flink_ from official flink image, change if necessary
      volumes:
        - name: flink-config-volume
          configMap:
            name: flink-config
            items:
              - key: flink-conf.yaml
                path: flink-conf.yaml
              - key: log4j-console.properties
                path: log4j-console.properties
        - name: job-artifacts-volume
          hostPath:
            path: /mnt/nfsdata/flink/application/job-artifacts

【温馨提示】注意这里的挂载/mnt/bigdata/flink/usrlib,最好这里使用共享目录。

taskmanager-job-deployment.yaml
apiVersion: apps/v1
kind: Deployment
metadata:
  name: flink-taskmanager
spec:
  replicas: 2
  selector:
    matchLabels:
      app: flink
      component: taskmanager
  template:
    metadata:
      labels:
        app: flink
        component: taskmanager
    spec:
      containers:
      - name: taskmanager
        image: myharbor.com/bigdata/flink-centos-admin:1.14.6-scala_2.12
        env:
        args: ["taskmanager"]
        ports:
        - containerPort: 6122
          name: rpc
        - containerPort: 6125
          name: query-state
        livenessProbe:
          tcpSocket:
            port: 6122
          initialDelaySeconds: 30
          periodSeconds: 60
        volumeMounts:
        - name: flink-config-volume
          mountPath: /opt/apache/flink-1.14.6/conf
        - name: job-artifacts-volume
          mountPath: /opt/apache/flink-1.14.6/usrlib
        securityContext:
          runAsUser: 9999  # refers to user _flink_ from official flink image, change if necessary
      volumes:
      - name: flink-config-volume
        configMap:
          name: flink-config
          items:
          - key: flink-conf.yaml
            path: flink-conf.yaml
          - key: log4j-console.properties
            path: log4j-console.properties
      - name: job-artifacts-volume
        hostPath:
          path: /mnt/nfsdata/flink/application/job-artifacts

【4】创建flink集群并提交任务

kubectl create ns flink
# Configuration and service definition
kubectl create -f flink-configuration-configmap.yaml -n flink

# service
kubectl create -f jobmanager-service.yaml -n flink
kubectl create -f jobmanager-rest-service.yaml -n flink
kubectl create -f taskmanager-query-state-service.yaml -n flink

# Create the deployments for the cluster
kubectl create -f  jobmanager-application-non-ha.yaml -n flink
kubectl create -f  taskmanager-job-deployment.yaml -n flink

查看

kubectl get pods,svc -n flink

Flink on k8s 讲解与实战操作_第17张图片

【5】删除flink集群

kubectl delete -f flink-configuration-configmap.yaml -n flink
kubectl delete -f jobmanager-service.yaml -n flink
kubectl delete -f jobmanager-rest-service.yaml -n flink
kubectl delete -f taskmanager-query-state-service.yaml -n flink
kubectl delete -f jobmanager-application-non-ha.yaml -n flink
kubectl delete -f taskmanager-job-deployment.yaml -n flink

kubectl delete ns flink --force

【6】查看

kubectl get pods,svc -n flink
kubectl exec -it flink-taskmanager-54cb7fc57c-g484q -n flink -- bash

Flink on k8s 讲解与实战操作_第18张图片

链接:https://www.cnblogs.com/liugp/p/16755095.html

你可能感兴趣的:(云计算,智能运维,flink,kubernetes,scala)