当前位置： 首页 > news >正文

k8s之flink的几种创建方式

news 来源：原创 2024/5/8 19:40:23

在此之前需要部署一下私人docker仓库，教程搭建 Docker 镜像仓库

注意：每台节点的daemon.json都需要配置"insecure-registries": ["http://主机IP:8080"] 并重启

一、session 模式

Session 模式是指在 Kubernetes 上启动一个共享的 Flink 集群（由 JobManager 和多个 TaskManagers 组成），然后多个 Flink 作业可以提交到这个共享集群上运行。这个模式下的集群会长期运行，直到用户手动停止它。这种模式适合多个作业需要频繁启动和停止，且对集群资源的利用率要求较高的场景。


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

1.1 Native Kubernetes 模式

Flink 的 Native Kubernetes 模式允许用户将 Apache Flink 无缝集成至 Kubernetes 环境中，实现在 Kubernetes 上运行 Flink 作业和应用程序。这种模式的主要优点是 Flink 能够利用 Kubernetes 提供的资源编排和管理能力，简化 Flink 集群的部署和管理。

在 Native Kubernetes 模式下，Flink 集群的部署和管理是通过 Flink 的 Kubernetes Operator 或者是直接使用 kubectl 命令行工具来完成的。Flink 的每个组件都被作为 Kubernetes 资源（如Pods, Services等）来管理。

1.1.1 构建镜像 Dockerfile

1.创建dockerfileFROM flink:1.16.2
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-82.开始构建镜像
docker build -t  192.168.20.62:2333/bigdata/flink-session:1.16.23.上传镜像
docker push 192.168.20.62:2333/bigdata/flink-session:1.16.2

1.1.2 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

1.1.3 创建 flink 集群

./bin/kubernetes-session.sh \
-Dkubernetes.cluster-id=my-first-flink-cluster  \
-Dkubernetes.container.image=192.168.20.62:2333/bigdata/flink-session:1.16.2 \
-Dkubernetes.namespace=flink \
-Dkubernetes.jobmanager.service-account=flink-service-account \
-Dkubernetes.rest-service.exposed.type=NodePort

1.1.4 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 \
-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 \

1.1.5 删除 flink 集群

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

1.2 Standalone 模式

Standalone 模式通常指的是在 Kubernetes 集群上运行 Flink 的一个单独集群环境，但它不是专门为 Kubernetes 设计的。在 Kubernetes 上使用 Standalone 模式意味着你将手动设置 Flink 集群（包括 JobManager 和 TaskManagers），而不是通过 Kubernetes Operator 或者其他 Kubernetes 原生的资源调度和管理机制。换句话说，在这个模式下，Flink 集群的各个组件（JobManager和TaskManagers）运行在 Kubernetes Pod 中，但是它们的生命周期管理并不是通过 Kubernetes 原生的支持来实现的，而是类似于在任何其他环境中部署 Flink 的传统方式。

1.2.1 创建docker-entrypoint.sh脚本

#!/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 [ $(id -u) != 0 ]; then# Don't need to drop privs if EUID != 0returnelif [ -x /sbin/su-exec ]; then# Alpineecho su-exec adminelse# Othersecho gosu adminfi
}copy_plugins_if_required() {if [ -z "$ENABLE_BUILT_IN_PLUGINS" ]; thenreturn 0fiecho "Enabling required built-in plugins"for target_plugin in $(echo "$ENABLE_BUILT_IN_PLUGINS" | tr ';' ' '); doecho "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}" ]; thenecho "Plugin ${target_plugin} does not exist. Exiting."exit 1elseln -fs "${FLINK_HOME}/opt/${target_plugin}" "${FLINK_HOME}/plugins/${plugin_name}"echo "Successfully enabled ${target_plugin}"fidone
}set_config_option() {local option=$1local value=$2# escape periods for usage in regular expressionslocal escaped_option=$(echo ${option} | sed -e "s/\./\\\./g")# either override an existing entry, or append a new oneif grep -E "^${escaped_option}:.*" "${CONF_FILE}" > /dev/null; thensed -i -e "s/${escaped_option}:.*/$option: $value/g" "${CONF_FILE}"elseecho "${option}: ${value}" >> "${CONF_FILE}"fi
}prepare_configuration() {set_config_option jobmanager.rpc.address ${JOB_MANAGER_RPC_ADDRESS}set_config_option blob.server.port 6124set_config_option query.server.port 6125if [ -n "${TASK_MANAGER_NUMBER_OF_TASK_SLOTS}" ]; thenset_config_option taskmanager.numberOfTaskSlots ${TASK_MANAGER_NUMBER_OF_TASK_SLOTS}fiif [ -n "${FLINK_PROPERTIES}" ]; thenecho "${FLINK_PROPERTIES}" >> "${CONF_FILE}"fienvsubst < "${CONF_FILE}" > "${CONF_FILE}.tmp" && mv "${CONF_FILE}.tmp" "${CONF_FILE}"
}maybe_enable_jemalloc() {if [ "${DISABLE_JEMALLOC:-false}" == "false" ]; thenJEMALLOC_PATH="/usr/lib/$(uname -m)-linux-gnu/libjemalloc.so"JEMALLOC_FALLBACK="/usr/lib/x86_64-linux-gnu/libjemalloc.so"if [ -f "$JEMALLOC_PATH" ]; thenexport LD_PRELOAD=$LD_PRELOAD:$JEMALLOC_PATHelif [ -f "$JEMALLOC_FALLBACK" ]; thenexport LD_PRELOAD=$LD_PRELOAD:$JEMALLOC_FALLBACKelseif [ "$JEMALLOC_PATH" = "$JEMALLOC_FALLBACK" ]; thenMSG_PATH=$JEMALLOC_PATHelseMSG_PATH="$JEMALLOC_PATH and $JEMALLOC_FALLBACK"fiecho "WARNING: attempted to load jemalloc from $MSG_PATH but the library couldn't be found. glibc will be used instead."fifi
}maybe_enable_jemalloccopy_plugins_if_requiredprepare_configurationargs=("$@")
if [ "$1" = "help" ]; thenprintf "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 [ "$1" = "jobmanager" ]; thenargs=("${args[@]:1}")echo "Starting Job Manager"exec $(drop_privs_cmd) "$FLINK_HOME/bin/jobmanager.sh" start-foreground "${args[@]}"
elif [ "$1" = ${COMMAND_STANDALONE} ]; thenargs=("${args[@]:1}")echo "Starting Job Manager"exec $(drop_privs_cmd) "$FLINK_HOME/bin/standalone-job.sh" start-foreground "${args[@]}"
elif [ "$1" = ${COMMAND_HISTORY_SERVER} ]; thenargs=("${args[@]:1}")echo "Starting History Server"exec $(drop_privs_cmd) "$FLINK_HOME/bin/historyserver.sh" start-foreground "${args[@]}"
elif [ "$1" = "taskmanager" ]; thenargs=("${args[@]:1}")echo "Starting Task Manager"exec $(drop_privs_cmd) "$FLINK_HOME/bin/taskmanager.sh" start-foreground "${args[@]}"
fiargs=("${args[@]}")# Running command in pass-through mode
exec $(drop_privs_cmd) "${args[@]}"

1.2.2 编排 Dockerfile

FROM centos:7.9.2009USER 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/timezoneRUN mkdir -p /opt/apacheADD jdk-8u231-linux-x64.tar.gz /opt/apache/ADD flink-1.16.2-bin-scala_2.12.tgz  /opt/apache/ENV FLINK_HOME /opt/apache/flink-1.16.2
ENV JAVA_HOME /opt/apache/jdk1.8.0_231
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.shRUN groupadd --system --gid=9999 admin && useradd --system --home-dir $FLINK_HOME --uid=9999 --gid=admin adminRUN 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"]

1.2.3 开始构建镜像

docker build -t 192.168.20.62:2333/bigdata/flink-centos-admin:1.16.2 . --no-cache# 上传镜像
docker push 192.168.20.62:2333/bigdata/flink-centos-admin:1.16.2

1.2.4 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

1.2.5 编排 yaml 文件

1.2.5.1  flink-configuration-configmap.yaml
apiVersion: v1
kind: ConfigMap
metadata:name: flink-configlabels:app: flink
data:flink-conf.yaml: |+jobmanager.rpc.address: flink-jobmanagertaskmanager.numberOfTaskSlots: 2blob.server.port: 6124jobmanager.rpc.port: 6123taskmanager.rpc.port: 6122queryable-state.proxy.ports: 6125jobmanager.memory.process.size: 3200mtaskmanager.memory.process.size: 2728mtaskmanager.memory.flink.size: 2280mparallelism.default: 2log4j-console.properties: |+# This affects logging for both user code and FlinkrootLogger.level = INFOrootLogger.appenderRef.console.ref = ConsoleAppenderrootLogger.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 = akkalogger.akka.level = INFOlogger.kafka.name= org.apache.kafkalogger.kafka.level = INFOlogger.hadoop.name = org.apache.hadooplogger.hadoop.level = INFOlogger.zookeeper.name = org.apache.zookeeperlogger.zookeeper.level = INFO# Log all infos to the consoleappender.console.name = ConsoleAppenderappender.console.type = CONSOLEappender.console.layout.type = PatternLayoutappender.console.layout.pattern = %d{yyyy-MM-dd HH:mm:ss,SSS} %-5p %-60c %x - %m%n# Log all infos in the given rolling fileappender.rolling.name = RollingFileAppenderappender.rolling.type = RollingFileappender.rolling.append = falseappender.rolling.fileName = ${sys:log.file}appender.rolling.filePattern = ${sys:log.file}.%iappender.rolling.layout.type = PatternLayoutappender.rolling.layout.pattern = %d{yyyy-MM-dd HH:mm:ss,SSS} %-5p %-60c %x - %m%nappender.rolling.policies.type = Policiesappender.rolling.policies.size.type = SizeBasedTriggeringPolicyappender.rolling.policies.size.size=100MBappender.rolling.strategy.type = DefaultRolloverStrategyappender.rolling.strategy.max = 10# Suppress the irrelevant (wrong) warnings from the Netty channel handlerlogger.netty.name = org.jboss.netty.channel.DefaultChannelPipelinelogger.netty.level = OFF
1.2.5.2 jobmanager-service.yaml
apiVersion: v1
kind: Service
metadata:name: flink-jobmanager
spec:type: ClusterIPports:- name: rpcport: 6123- name: blob-serverport: 6124- name: webuiport: 8081selector:app: flinkcomponent: jobmanager
1.2.5.3 jobmanager-rest-service.yaml

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: NodePortports:- name: restport: 8081targetPort: 8081nodePort: 30081selector:app: flinkcomponent: jobmanager
1.2.5.4 taskmanager-query-state-service.yaml

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: NodePortports:- name: query-stateport: 6125targetPort: 6125nodePort: 30025selector:app: flinkcomponent: taskmanager
1.2.5.5  jobmanager-session-deployment-non-ha.yaml
apiVersion: apps/v1
kind: Deployment
metadata:name: flink-jobmanager
spec:replicas: 1selector:matchLabels:app: flinkcomponent: jobmanagertemplate:metadata:labels:app: flinkcomponent: jobmanagerspec:containers:- name: jobmanagerimage: 192.168.20.62:2333/bigdata/flink-centos-admin:1.16.2args: ["jobmanager"]ports:- containerPort: 6123name: rpc- containerPort: 6124name: blob-server- containerPort: 8081name: webuilivenessProbe:tcpSocket:port: 6123initialDelaySeconds: 30periodSeconds: 60volumeMounts:- name: flink-config-volumemountPath: /opt/apache/flink-1.16.2/conf/securityContext:runAsUser: 9999  # refers to user _flink_ from official flink image, change if necessaryvolumes:- name: flink-config-volumeconfigMap:name: flink-configitems:- key: flink-conf.yamlpath: flink-conf.yaml- key: log4j-console.propertiespath: log4j-console.properties
1.2.5.6  taskmanager-session-deployment.yaml
apiVersion: apps/v1
kind: Deployment
metadata:name: flink-taskmanager
spec:replicas: 2selector:matchLabels:app: flinkcomponent: taskmanagertemplate:metadata:labels:app: flinkcomponent: taskmanagerspec:containers:- name: taskmanagerimage: 192.168.20.62:2333/bigdata/flink-centos-admin:1.16.2args: ["taskmanager"]ports:- containerPort: 6122name: rpc- containerPort: 6125name: query-statelivenessProbe:tcpSocket:port: 6122initialDelaySeconds: 30periodSeconds: 60volumeMounts:- name: flink-config-volumemountPath: /opt/apache/flink-1.16.2/conf/securityContext:runAsUser: 9999  # refers to user _flink_ from official flink image, change if necessaryvolumes:- name: flink-config-volumeconfigMap:name: flink-configitems:- key: flink-conf.yamlpath: flink-conf.yaml- key: log4j-console.propertiespath: log4j-console.properties

1.2.6 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

1.2.7 提交任务

./bin/flink run -m 192.168.20.62:30081 ./examples/streaming/TopSpeedWindowing.jar

1.2.8 删除集群

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

二 、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

2.1 Native Kubernetes 模式（常用）

2.1.1 构建镜像 Dockerfile

FROM flink:1.16.2
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   ./flink-1.16.2/examples/streaming/TopSpeedWindowing.jar /opt/flink/usrlib/开始构建镜像docker build -t 192.168.20.62:2333/bigdata/flink-application:1.16.2 . --no-cache
docker push  192.168.20.62:2333/bigdata/flink-application:1.16.2

2.1.2 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

2.1.3 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=192.168.20.62:2333/bigdata/flink-application:1.16.2 \-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

local是application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。

2.1.4 删除 flink 集群

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

2.2 Standalone 模式

在此之前需要使用nfs设置一个共享目录，配置文件中设置共享目录是/mnt/bigdata/flink/usrlib

NFS共享存储服务

2.2.1 创建启动脚本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 [ $(id -u) != 0 ]; then# Don't need to drop privs if EUID != 0returnelif [ -x /sbin/su-exec ]; then# Alpineecho su-exec adminelse# Othersecho gosu adminfi
}copy_plugins_if_required() {if [ -z "$ENABLE_BUILT_IN_PLUGINS" ]; thenreturn 0fiecho "Enabling required built-in plugins"for target_plugin in $(echo "$ENABLE_BUILT_IN_PLUGINS" | tr ';' ' '); doecho "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}" ]; thenecho "Plugin ${target_plugin} does not exist. Exiting."exit 1elseln -fs "${FLINK_HOME}/opt/${target_plugin}" "${FLINK_HOME}/plugins/${plugin_name}"echo "Successfully enabled ${target_plugin}"fidone
}set_config_option() {local option=$1local value=$2# escape periods for usage in regular expressionslocal escaped_option=$(echo ${option} | sed -e "s/\./\\\./g")# either override an existing entry, or append a new oneif grep -E "^${escaped_option}:.*" "${CONF_FILE}" > /dev/null; thensed -i -e "s/${escaped_option}:.*/$option: $value/g" "${CONF_FILE}"elseecho "${option}: ${value}" >> "${CONF_FILE}"fi
}prepare_configuration() {set_config_option jobmanager.rpc.address ${JOB_MANAGER_RPC_ADDRESS}set_config_option blob.server.port 6124set_config_option query.server.port 6125if [ -n "${TASK_MANAGER_NUMBER_OF_TASK_SLOTS}" ]; thenset_config_option taskmanager.numberOfTaskSlots ${TASK_MANAGER_NUMBER_OF_TASK_SLOTS}fiif [ -n "${FLINK_PROPERTIES}" ]; thenecho "${FLINK_PROPERTIES}" >> "${CONF_FILE}"fienvsubst < "${CONF_FILE}" > "${CONF_FILE}.tmp" && mv "${CONF_FILE}.tmp" "${CONF_FILE}"
}maybe_enable_jemalloc() {if [ "${DISABLE_JEMALLOC:-false}" == "false" ]; thenJEMALLOC_PATH="/usr/lib/$(uname -m)-linux-gnu/libjemalloc.so"JEMALLOC_FALLBACK="/usr/lib/x86_64-linux-gnu/libjemalloc.so"if [ -f "$JEMALLOC_PATH" ]; thenexport LD_PRELOAD=$LD_PRELOAD:$JEMALLOC_PATHelif [ -f "$JEMALLOC_FALLBACK" ]; thenexport LD_PRELOAD=$LD_PRELOAD:$JEMALLOC_FALLBACKelseif [ "$JEMALLOC_PATH" = "$JEMALLOC_FALLBACK" ]; thenMSG_PATH=$JEMALLOC_PATHelseMSG_PATH="$JEMALLOC_PATH and $JEMALLOC_FALLBACK"fiecho "WARNING: attempted to load jemalloc from $MSG_PATH but the library couldn't be found. glibc will be used instead."fifi
}maybe_enable_jemalloccopy_plugins_if_requiredprepare_configurationargs=("$@")
if [ "$1" = "help" ]; thenprintf "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 [ "$1" = "jobmanager" ]; thenargs=("${args[@]:1}")echo "Starting Job Manager"exec $(drop_privs_cmd) "$FLINK_HOME/bin/jobmanager.sh" start-foreground "${args[@]}"
elif [ "$1" = ${COMMAND_STANDALONE} ]; thenargs=("${args[@]:1}")echo "Starting Job Manager"exec $(drop_privs_cmd) "$FLINK_HOME/bin/standalone-job.sh" start-foreground "${args[@]}"
elif [ "$1" = ${COMMAND_HISTORY_SERVER} ]; thenargs=("${args[@]:1}")echo "Starting History Server"exec $(drop_privs_cmd) "$FLINK_HOME/bin/historyserver.sh" start-foreground "${args[@]}"
elif [ "$1" = "taskmanager" ]; thenargs=("${args[@]:1}")echo "Starting Task Manager"exec $(drop_privs_cmd) "$FLINK_HOME/bin/taskmanager.sh" start-foreground "${args[@]}"
fiargs=("${args[@]}")# Running command in pass-through mode
exec $(drop_privs_cmd) "${args[@]}"

2.2.2 编排Dockerfile

FROM centos:7.9.2009USER 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/timezoneRUN mkdir -p /opt/apacheADD jdk-8u231-linux-x64.tar.gz /opt/apache/ADD flink-1.16.2-bin-scala_2.12.tgz  /opt/apache/ENV FLINK_HOME /opt/apache/flink-1.16.2
ENV JAVA_HOME /opt/apache/jdk1.8.0_231
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/COPY /RUN groupadd --system --gid=9999 admin && useradd --system --home-dir $FLINK_HOME --uid=9999 --gid=admin adminRUN chown -R admin:admin /opt/apache
RUN chmod +x /opt/apache/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 192.168.20.62:2333/bigdata/flink-centos-admin-application:1.16.2 . --no-cache
docker push 192.168.20.62:2333/bigdata/flink-centos-admin-application:1.16.2

2.2.3 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

2.2.4 flink-configuration-configmap.yaml

apiVersion: v1
kind: ConfigMap
metadata:name: flink-configlabels:app: flink
data:flink-conf.yaml: |+jobmanager.rpc.address: flink-jobmanagertaskmanager.numberOfTaskSlots: 2blob.server.port: 6124jobmanager.rpc.port: 6123taskmanager.rpc.port: 6122queryable-state.proxy.ports: 6125jobmanager.memory.process.size: 3200mtaskmanager.memory.process.size: 2728mtaskmanager.memory.flink.size: 2280mparallelism.default: 2log4j-console.properties: |+# This affects logging for both user code and FlinkrootLogger.level = INFOrootLogger.appenderRef.console.ref = ConsoleAppenderrootLogger.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 = akkalogger.akka.level = INFOlogger.kafka.name= org.apache.kafkalogger.kafka.level = INFOlogger.hadoop.name = org.apache.hadooplogger.hadoop.level = INFOlogger.zookeeper.name = org.apache.zookeeperlogger.zookeeper.level = INFO# Log all infos to the consoleappender.console.name = ConsoleAppenderappender.console.type = CONSOLEappender.console.layout.type = PatternLayoutappender.console.layout.pattern = %d{yyyy-MM-dd HH:mm:ss,SSS} %-5p %-60c %x - %m%n# Log all infos in the given rolling fileappender.rolling.name = RollingFileAppenderappender.rolling.type = RollingFileappender.rolling.append = falseappender.rolling.fileName = ${sys:log.file}appender.rolling.filePattern = ${sys:log.file}.%iappender.rolling.layout.type = PatternLayoutappender.rolling.layout.pattern = %d{yyyy-MM-dd HH:mm:ss,SSS} %-5p %-60c %x - %m%nappender.rolling.policies.type = Policiesappender.rolling.policies.size.type = SizeBasedTriggeringPolicyappender.rolling.policies.size.size=100MBappender.rolling.strategy.type = DefaultRolloverStrategyappender.rolling.strategy.max = 10# Suppress the irrelevant (wrong) warnings from the Netty channel handlerlogger.netty.name = org.jboss.netty.channel.DefaultChannelPipelinelogger.netty.level = OFF

2.2.5 jobmanager-service.yaml

apiVersion: v1
kind: Service
metadata:name: flink-jobmanager
spec:type: ClusterIPports:- name: rpcport: 6123- name: blob-serverport: 6124- name: webuiport: 8081selector:app: flinkcomponent: jobmanager

2.2.6  jobmanager-rest-service.yaml

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: NodePortports:- name: restport: 8081targetPort: 8081nodePort: 30081selector:app: flinkcomponent: jobmanager

2.2.7 taskmanager-query-state-service.yaml

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: NodePortports:- name: query-stateport: 6125targetPort: 6125nodePort: 30025selector:app: flinkcomponent: taskmanager

2.2.8 jobmanager-application-non-ha.yaml

apiVersion: batch/v1
kind: Job
metadata:name: flink-jobmanager
spec:template:metadata:labels:app: flinkcomponent: jobmanagerspec:restartPolicy: OnFailurecontainers:- name: jobmanagerimage: 192.168.20.62:2333/bigdata/flink-centos-admin-application:1.16.2env:args: ["standalone-job", "--job-classname", "org.apache.flink.streaming.examples.windowing.TopSpeedWindowing","--output","/tmp/result"]ports:- containerPort: 6123name: rpc- containerPort: 6124name: blob-server- containerPort: 8081name: webuilivenessProbe:tcpSocket:port: 6123initialDelaySeconds: 30periodSeconds: 60volumeMounts:- name: flink-config-volumemountPath: /opt/apache/flink-1.16.2/conf- name: job-artifacts-volumemountPath: /opt/apache/flink-1.16.2/usrlibsecurityContext:runAsUser: 9999  # refers to user _flink_ from official flink image, change if necessaryvolumes:- name: flink-config-volumeconfigMap:name: flink-configitems:- key: flink-conf.yamlpath: flink-conf.yaml- key: log4j-console.propertiespath: log4j-console.properties- name: job-artifacts-volumehostPath:path: /mnt/bigdata/flink/usrlib

2.2.9 taskmanager-job-deployment.yaml

apiVersion: apps/v1
kind: Deployment
metadata:name: flink-taskmanager
spec:replicas: 2selector:matchLabels:app: flinkcomponent: taskmanagertemplate:metadata:labels:app: flinkcomponent: taskmanagerspec:containers:- name: taskmanagerimage: 192.168.20.62:2333/bigdata/flink-centos-admin-application:1.16.2env:args: ["taskmanager"]ports:- containerPort: 6122name: rpc- containerPort: 6125name: query-statelivenessProbe:tcpSocket:port: 6122initialDelaySeconds: 30periodSeconds: 60volumeMounts:- name: flink-config-volumemountPath: /opt/apache/flink-1.16.2/conf- name: job-artifacts-volumemountPath: /opt/apache/flink-1.16.2/usrlibsecurityContext:runAsUser: 9999  # refers to user _flink_ from official flink image, change if necessaryvolumes:- name: flink-config-volumeconfigMap:name: flink-configitems:- key: flink-conf.yamlpath: flink-conf.yaml- key: log4j-console.propertiespath: log4j-console.properties- name: job-artifacts-volumehostPath:path: /mnt/bigdata/flink/usrlib

2.2.10 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

2.2.11 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 flinkkubectl delete ns flink --force

相关文章：

  • ASP.NET Core中实现个人资料上传图片功能
  • 嵌入式Linux C语言介绍
  • 测试分类篇
  • 微服务-理论(CAP,一致性协议)
  • opencv期末练习题（3）附带解析
  • Spring之AOP
  • 25计算机专业考研经验贴之准备篇
  • SpringCloud系列篇：核心组件之注册中心组件
  • 大津法（OTSU）点云强度信息分割
  • 安装jupyter notebook，jupyter notebook的简单使用
  • 橘子学K8S03之容器的理解
  • Android aar打包集成问题处理合集
  • 2023年12 月电子学会Python等级考试试卷（四级）答案解析
  • 【CSS】文字描边的三种实现方式
  • 微信小程序封装vant 下拉框select 单选组件
  • [微信小程序] 使用ES6特性Class后出现编译异常
  • 2017 年终总结 —— 在路上
  • github指令
  • JavaScript异步流程控制的前世今生
  • js算法-归并排序（merge_sort）
  • magento 货币换算
  • mockjs让前端开发独立于后端
  • nodejs调试方法
  • 阿里云爬虫风险管理产品商业化，为云端流量保驾护航
  • 从@property说起（二）当我们写下@property (nonatomic, weak) id obj时，我们究竟写了什么...
  • 如何胜任知名企业的商业数据分析师？
  • 一起来学SpringBoot | 第三篇：SpringBoot日志配置
  • 移动端唤起键盘时取消position:fixed定位
  • ​直流电和交流电有什么区别为什么这个时候又要变成直流电呢？交流转换到直流（整流器）直流变交流（逆变器）​
  • #微信小程序（布局、渲染层基础知识）
  • （C语言）共用体union的用法举例
  • (Python) SOAP Web Service (HTTP POST)
  • （二）丶RabbitMQ的六大核心
  • (附源码)springboot人体健康检测微信小程序 毕业设计 012142
  • （论文阅读32/100）Flowing convnets for human pose estimation in videos
  • (三)mysql_MYSQL(三)
  • （淘宝无限适配）手机端rem布局详解（转载非原创）
  • （一）Thymeleaf用法——Thymeleaf简介
  • .net mvc部分视图
  • .NET 中使用 Mutex 进行跨越进程边界的同步
  • .NET/C# 项目如何优雅地设置条件编译符号？
  • @DateTimeFormat 和 @JsonFormat 注解详解
  • [ 英语 ] 马斯克抱水槽“入主”推特总部中那句 Let that sink in 到底是什么梗？
  • [Asp.net MVC]Bundle合并，压缩js、css文件
  • [AutoSar]BSW_Com07 CAN报文接收流程的函数调用
  • [BROADCASTING]tensor的扩散机制
  • [BUUCTF]-PWN:[极客大挑战 2019]Not Bad解析
  • [C#基础]说说lock到底锁谁？
  • [C++基础]-入门知识
  • [Codeforces] probabilities (R1600) Part.1
  • [DAX] MAX函数 | MAXX函数
  • [Django 0-1] Core.Handlers 模块
  • [E单调栈] lc2487. 从链表中移除节点(单调栈+递归+反转链表+多思路)
  • [Gamma]阶段测试报告
  • [HackMyVM]靶场 VivifyTech