使用Jenkins和单个模板部署多个Kubernetes组件

前言

在持续集成和部署中，我们通常需要部署多个实例或组件到Kubernetes集群中。通过Jenkins的管道脚本，我们可以自动化这个过程。在本文中，我将演示如何使用Jenkins Pipeline及单个YAML模板文件（.tpl）来部署多个类似的Kubernetes组件，而不需要为每个组件提供单独的模板文件。

问题背景

参照：Jenkins Pipeline 脚本优化实践：从繁琐到简洁 批量生成 Kubernetes 部署模板：从 1 到20顺序模板

pipeline {agent none // Use none at the top level, each stage will define its own agent.environment {REGISTRY = "xxxx/master-metaspace"KUBE_CONFIG = "--namespace=master-metaspace --context=master"KUBE_YAML_PATH = "/home/jenkins/workspace/yaml/master-metaspace"// Assume that 'data' is defined elsewhere or injected as a parameter.BASE_WORKSPACE = "xxxxxxx" // 定义一个基础工作空间路径}stages {stage("GetCode") {agent { label "build01" }steps {script {checkout scm: [$class: 'GitSCM',branches: [[name: env.branchName]],extensions: [[$class: 'CloneOption', depth: 1, noTags: false, shallow: true]],userRemoteConfigs: [[credentialsId: 'xxxx', url: env.gitHttpURL]]]}}}stage("Docker Builds") {parallel {stage('Build dataloader-game-ucenter') {agent { label "build01" }when { environment name: 'dataloader', value: 'true' }steps {buildAndPushDockerImage("dataloader-game-ucenter", env.data, env.BASE_WORKSPACE)}}stage('Build datawriter-game-ucenter') {agent { label "build01" }when { environment name: 'datawriter', value: 'true' }steps {buildAndPushDockerImage("datawriter-game-ucenter", env.data, env.BASE_WORKSPACE)}}stage('Build game-ucenter') {agent { label "build01" }when { environment name: 'game-ucenter', value: 'true' }steps {buildAndPushDockerImage("game-ucenter", env.data, env.BASE_WORKSPACE)}}}}stage('Development Deployment') {parallel {stage("Deploy datawriter-game-ucenter") {when { environment name: 'datawriter-game-ucenter', value: 'true' }agent { label  "huaweiyun-xx" }steps {deployToKubernetes("datawriter-game-ucenter")}}stage("Deploy dataloader-game-ucenter") {when { environment name: 'dataloader', value: 'true' }agent { label  "huaweiyun-xx" }steps {deployToKubernetes("dataloader-game-ucenter")}}stage("Deploy game-ucenter") {when { environment name: 'game-ucenter', value: 'true' }agent { label  "huaweiyun-xx" }steps {deployToKubernetes("game-ucenter-1")deployToKubernetes("game-ucenter-2")deployToKubernetes("game-ucenter-3")deployToKubernetes("game-ucenter-4")............................}}}}}
}// Define methods outside pipeline to avoid repetitiondef buildAndPushDockerImage(String imageName, String tag, String workspacePath) {sh "cd ${workspacePath} && echo 'Current directory: \$(pwd)'" // 使用基础工作空间变量sh "cd ${workspacePath}/${imageName}&& docker build --build-arg NODE_ENV=$imageName -t $REGISTRY/$imageName:$tag ."withCredentials([usernamePassword(credentialsId: 'xxxxx', passwordVariable: 'dockerPassword', usernameVariable: 'dockerUser')]) {sh "docker login -u $dockerUser -p $dockerPassword $REGISTRY"sh "docker push $REGISTRY/$imageName:$tag"}
}def deployToKubernetes(String kubernetesComponent) {String templateFile = "${KUBE_YAML_PATH}/${kubernetesComponent}.tpl"String outputFile = "${KUBE_YAML_PATH}/${kubernetesComponent}.yaml"sh "sed -e 's/{data}/$data/g' $templateFile > $outputFile"sh "sudo kubectl apply -f $outputFile $KUBE_CONFIG"
}

默认jenkins pipeline如上，我们有多个相似的游戏用户中心服务game-ucenter-*运行在Kubernetes集群中，它们都使用非常相似的Kubernetes YAML配置文件，配置文件之间的差异主要是一些标识符的不同（例如，服务的序号）。在传统的做法中，维护一系列几乎一样的模板文件（如game-ucenter-1.tpl, game-ucenter-2.tpl 等）将非常低效且易出错。

为了精简流程和提高效率，我们需要一个方法来通过单一模板生成多个配置文件，并由此部署多个不同的服务实例。

解决方案

使用Jenkins Pipeline中的sed命令和循环结构，我们可以从单一模板生成多个Kubernetes配置文件，并相应地部署每个服务实例。参照generate_templates.sh脚本

#!/bin/bash# Define the name of the template file.
TEMPLATE_FILE="game-ucenter.tpl"# Check if the template file exists.
if [ ! -f "$TEMPLATE_FILE" ]; thenecho "Template file $TEMPLATE_FILE does not exist."exit 1
fi# Loop to create files from game-ucenter-2 to pvp-game-20 based on the template.
for i in $(seq 1 20); do# Define the name of the new file.NEW_FILE="game-ucenter-${i}.yaml"# Copy the template to the new file.cp $TEMPLATE_FILE $NEW_FILE# Use 'sed' to replace 'game-ucenter-1' with 'game-ucenter-N' and save inline (-i option).sed -i "s/game-ucenter/game-ucenter-${i}/g" $NEW_FILEecho "Created file: $NEW_FILE"
doneecho "All files created successfully."

步骤 1: 定义Jenkins Pipeline

在我们的Jenkins脚本中，我们首先定义了基础环境变量和两个函数：buildAndPushDockerImage 和 deployToKubernetes。这些函数将用于构建Docker镜像并部署到Kubernetes

def buildAndPushDockerImage(String imageName, String tag, String workspacePath) {sh "cd ${workspacePath} && echo 'Current directory: \$(pwd)'" // 使用基础工作空间变量sh "cd ${workspacePath}/${imageName}&& docker build --build-arg NODE_ENV=$imageName -t $REGISTRY/$imageName:$tag ."withCredentials([usernamePassword(credentialsId: 'xxx', passwordVariable: 'dockerPassword', usernameVariable: 'dockerUser')]) {sh "docker login -u $dockerUser -p $dockerPassword $REGISTRY"sh "docker push $REGISTRY/$imageName:$tag"}
}def deployToKubernetes(String kubernetesComponent) {String templateFile = "${KUBE_YAML_PATH}/${kubernetesComponent}.tpl"String outputFile = "${KUBE_YAML_PATH}/${kubernetesComponent}.yaml"sh "sed -e 's/{data}/$data/g' $templateFile > $outputFile"sh "sudo kubectl apply -f $outputFile $KUBE_CONFIG"
}

步骤 2: 修改deployToKubernetes函数

接下来，我们需要修改deployToKubernetes函数，以便它能够接受组件名称，并使用单一模板文件创建具体的配置文件。

def deployToKubernetes(String kubernetesComponent, int instance=1, int totalInstances=1) {// 检查实例值if (instance < 1) {error("实例数必须大于0")}// 根据 instance 的值来定义资源的名称和文件名String nameSuffix = totalInstances > 1 ? "-${instance}" : "" // 总是添加后缀，除非只有一个实例String outputFileName = "${kubernetesComponent}${nameSuffix}.yaml"String templateFile = "${KUBE_YAML_PATH}/${kubernetesComponent}.tpl"String outputFile = "${KUBE_YAML_PATH}/${outputFileName}"String nameReplacement = "${kubernetesComponent}${nameSuffix}"sh """cat "${templateFile}" \| sed 's/{data}/${data}/g' \| sed 's/name: ${kubernetesComponent}/name: ${nameReplacement}/g' \| sed 's/app: ${kubernetesComponent}/app: ${nameReplacement}/g' \> "${outputFile}""""// 使用 KUBE_CONFIG 应用 Kubernetes 配置sh "kubectl apply -f ${outputFile} ${KUBE_CONFIG}"
}

对于单实例的业务，例如Deploy dataloader-game-ucenter，我们不需要传递实例编号。

stage("Deploy dataloader-game-ucenter") {when { environment name: 'dataloader', value: 'true' }agent { label  "huaweiyun-xx" }steps {deployToKubernetes("dataloader-game-ucenter")}
}

对于多实例。我这里生成 规则优点强迫症了。如果多实例我生成的规则要求符合game-ucenter-1,game-ucenter-2,game-ucenter-3…顺序，当单个实例的时候则保持原来的不加标签：

                stage("Deploy game-ucenter") {when { environment name: 'game-ucenter', value: 'true' }agent { label  "k8s-node-06" }steps {script {int instances = 2 // 假设我们有2个实例for (int i = 1; i <= instances; i++) {def componentName = "game-ucenter"deployToKubernetes("game-ucenter", i, instances)}}}}

步骤 3: 准备模板文件

我们的模板文件game-ucenter.tpl将包含通用的Kubernetes服务或部署定义，使用占位符game-ucenter-1 game-ucenter-2来标识应该被替换的地方。

# game-ucenter-1.tpl (示例部分)
apiVersion: apps/v1
kind: Deployment
metadata:name: game-ucenter
spec:replicas: 1strategy:rollingUpdate:maxSurge: 1maxUnavailable: 0selector:matchLabels:app: game-ucentertemplate:metadata:labels:app: game-ucenterspec:containers:- name: game-ucenterimage: xxxx/xxx/game-ucenter:{data}envFrom:- configMapRef:name: deployports:- containerPort: 80resources:requests:memory: "4096M"cpu: "2000m"limits:memory: "4096M"cpu: "2000m" livenessProbe:httpGet:scheme: HTTPpath: /test.htmlport: 80initialDelaySeconds: 20periodSeconds: 120successThreshold: 1failureThreshold: 3readinessProbe:httpGet:scheme: HTTPpath: /test.htmlport: 80initialDelaySeconds: 20periodSeconds: 120imagePullSecrets:                                              - name: xxx
---apiVersion: v1
kind: Service
metadata:name: game-ucenterlabels:app: game-ucenter
spec:ports:- port: 80protocol: TCPtargetPort: 80selector:app: game-ucenter# ...

步骤 4: 执行Jenkins Pipeline

当Jenkins Pipeline运行到"Development Deployment"阶段时，它将循环创建和应用game-ucenter-1.yaml到game-ucenter-2.yaml的配置文件，从而部署2个game-ucenterdeployment服务实例。

并保证单个实例的原有命名规则：

通过这一方法，我们不再需要为每个服务实例维护一个单独的模板文件，而是可以通过一个模板文件和Jenkins Pipeline的自动化来简化服务部署工作。这样做不仅提升了效率，也降低了出错的风险。

注意：

以上代码和命令为示例性质，可能需要根据您具体的Jenkins环境和Kubernetes集群进行相应的调整。在生产环境中部署之前，请确保进行充分的测试。