2

DHorse(K8S的CICD平台)的实现原理 - tiandizhiguai

 10 months ago
source link: https://www.cnblogs.com/tiandizhiguai/p/17819241.html
Go to the source link to view the article. You can view the picture content, updated content and better typesetting reading experience. If the link is broken, please click the button below to view the snapshot at that time.
neoserver,ios ssh client

DHorse(K8S的CICD平台)的实现原理

首先,本篇文章所介绍的内容,已经有完整的实现,可以参考这里
在微服务、DevOps和云平台流行的当下,使用一个高效的持续集成工具也是一个非常重要的事情。虽然市面上目前已经存在了比较成熟的自动化构建工具,比如jekines,还有一些商业公司推出的自动化构建工具,但他们都不能够很好的和云环境相结合。那么究竟该如何实现一个简单、快速的基于云环境的自动化构建系统呢?我们首先以一个Springboot应用为例来介绍一下整体的发布流程,然后再来看看具体如何实现。发布的步骤大体如下:
1.首先从代码仓库下载代码,比如Gitlab、GitHub等;
2.接着是进行打包,比如使用Maven、Gradle等;
3.如果要使用k8s作为编排,还需要把步骤2产生的包制作成镜像,比如用Docker等;
4.上传步骤3的镜像到远程仓库,比如Harhor、DockerHub等;
5.最后,下载镜像并编写Deployment文件部署到k8s集群;
如图1所示:

在这里插入图片描述

图1

从以上步骤可以看出,发布过程中需要的工具和环境至少包括:代码仓库(Gitlab、GitHub等)、打包环境(Maven、Gradle等)、镜像制作(Docker等)、镜像仓库(Harbor、DockerHub等)、k8s集群等;此外,还包括发布系统自身的数据存储等。
可以看出,整个流程里依赖的环境很多,如果发布系统不能与这些环境解耦,那么要想实现一个安装简单、功能快速的系统没有那么容易。那么有没有合理的解决方案来实现与这些环境的解耦呢?答案是有的,下面就分别介绍。

操作代码仓库,一般系统提供的都有对应Restful API,以GitLab系统提供的Java客户端为例,如下代码:

<dependency>
	<groupId>org.gitlab4j</groupId>
	<artifactId>gitlab4j-api</artifactId>
	<version>4.17.0</version>
</dependency>

比如,我们想获取某个项目的分支列表,如下代码所示:

public List<Branch> branchList(CodeRepo codeRepo, BranchListParam param) {
	GitLabApi gitLabApi = gitLabApi(codeRepo);
	List<Branch> list = null;
	try {
		list = gitLabApi.getRepositoryApi().getBranches(param.getProjectIdOrPath(), param.getBranchName());
	} catch (GitLabApiException e) {
		LogUtils.throwException(logger, e, MessageCodeEnum.PROJECT_BRANCH_PAGE_FAILURE);
	} finally {
		gitLabApi.close();
	}
}

private GitLabApi gitLabApi(CodeRepo codeRepo) {
	GitLabApi gitLabApi = new GitLabApi(codeRepo.getUrl(), codeRepo.getAuthToken());
	gitLabApi.setRequestTimeout(1000, 5 * 1000);
	try {
		gitLabApi.getVersion();
	}catch(GitLabApiException e) {
		//如果token无效,则用账号登录
		if(e.getHttpStatus() == 401 && !StringUtils.isBlank(codeRepo.getAuthUser())) {
			gitLabApi = new GitLabApi(codeRepo.getUrl(), codeRepo.getAuthUser(), codeRepo.getAuthPassword());
			gitLabApi.setRequestTimeout(1000, 5 * 1000);
		}
	}
	
	return gitLabApi;
}

我们以Maven为例进行说明,一般情况下,我们使用Maven打包时,需要首先安装Maven环境,接着引入打包插件,然后使用mvn clean package命令就可以打包了。比如springboot自带插件:

<plugin>
	<groupId>org.springframework.boot</groupId>
	<artifactId>spring-boot-maven-plugin</artifactId>
	<version>2.5.6</version>
	<configuration>
		<classifier>execute</classifier>
		<mainClass>com.test.Application</mainClass>
	</configuration>
	<executions>
		<execution>
			<goals>
				<goal>repackage</goal>
			</goals>
		</execution>
	</executions>
</plugin>

再比如,通用的打包插件:

<plugin>
	<groupId>org.apache.maven.plugins</groupId>
	<artifactId>maven-assembly-plugin</artifactId>
	<version>3.8.2</version>
	<configuration>
		<appendAssemblyId>false</appendAssemblyId>
		<descriptors>
			<descriptor>src/main/resources/assemble.xml</descriptor>
		</descriptors>
		<outputDirectory>../target</outputDirectory>
	</configuration>
	<executions>
		<execution>
			<id>make-assembly</id>
			<phase>package</phase>
			<goals>
				<goal>single</goal>
			</goals>
		</execution>
	</executions>
</plugin>

等等。然后再通过运行mvn clean package命令进行打包。那么,在打包时如果要去除对maven环境的依赖,该如何实现呢?
可以使用嵌入式maven插件maven-embedder来实现。
具体可以这样来做,首先在平台项目里引入依赖,如下:

<dependency>
	<groupId>org.apache.maven</groupId>
	<artifactId>maven-embedder</artifactId>
	<version>3.8.1</version>
</dependency>
<dependency>
	<groupId>org.apache.maven</groupId>
	<artifactId>maven-compat</artifactId>
	<version>3.8.1</version>
</dependency>
<dependency>
	<groupId>org.apache.maven.resolver</groupId>
	<artifactId>maven-resolver-connector-basic</artifactId>
	<version>1.7.1</version>
</dependency>
<dependency>
	<groupId>org.apache.maven.resolver</groupId>
	<artifactId>maven-resolver-transport-http</artifactId>
	<version>1.7.1</version>
</dependency>

运行如下代码,就可以对项目进行打包了:

String[] commands = new String[] { "clean", "package", "-Dmaven.test.skip" };
String pomPath = "D:/hello/pom.xml";
MavenCli cli = new MavenCli();
try {
	cli.doMain(commands, pomPath, System.out, System.out);
} catch (Exception e) {
	e.printStackTrace();
}

但是,一般情况下,我们通过maven的settings文件还会做一些配置,比如配置工作目录、nexus私服地址、Jdk版本、编码方式等等,如下:

<?xml version="1.0" encoding="UTF-8"?>
<settings xmlns="http://maven.apache.org/SETTINGS/1.0.0"
	xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
	xsi:schemaLocation="http://maven.apache.org/SETTINGS/1.0.0 http://maven.apache.org/xsd/settings-1.0.0.xsd">
	<localRepository>C:/m2/repository</localRepository>
	<profiles>
		<profile>
			<id>myNexus</id>
			<repositories>
				<repository>
					<id>nexus</id>
					<name>nexus</name>
					<url>https://repo.maven.apache.org/maven2</url>
					<releases>
						<enabled>true</enabled>
					</releases>
					<snapshots>
						<enabled>true</enabled>
					</snapshots>
				</repository>
			</repositories>
			<pluginRepositories>
				<pluginRepository>
					<id>nexus</id>
					<name>nexus</name>
					<url>https://repo.maven.apache.org/maven2</url>
					<releases>
						<enabled>true</enabled>
					</releases>
					<snapshots>
						<enabled>true</enabled>
					</snapshots>
				</pluginRepository>
			</pluginRepositories>
		</profile>

		<profile>
			<id>java11</id>
			<activation>
				<activeByDefault>true</activeByDefault>
				<jdk>11</jdk>
			</activation>
			<properties>
				<maven.compiler.source>11</maven.compiler.source>
				<maven.compiler.target>11</maven.compiler.target>
				<maven.compiler.compilerVersion>11</maven.compiler.compilerVersion>
				<project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
				<project.build.outputEncoding>UTF-8</project.build.outputEncoding>
			</properties>
		</profile>
	</profiles>
	<activeProfiles>
		<activeProfile>myNexus</activeProfile>
	</activeProfiles>
</settings>

通过查看MavenCli类发现,doMain(CliRequest cliRequest)方法有比较丰富的参数,CliRequest的代码如下:

package org.apache.maven.cli;

public class CliRequest
{
    String[] args;

    CommandLine commandLine;

    ClassWorld classWorld;

    String workingDirectory;

    File multiModuleProjectDirectory;

    boolean debug;

    boolean quiet;

    boolean showErrors = true;

    Properties userProperties = new Properties();

    Properties systemProperties = new Properties();

    MavenExecutionRequest request;

    CliRequest( String[] args, ClassWorld classWorld )
    {
        this.args = args;
        this.classWorld = classWorld;
        this.request = new DefaultMavenExecutionRequest();
    }

    public String[] getArgs()
    {
        return args;
    }

    public CommandLine getCommandLine()
    {
        return commandLine;
    }

    public ClassWorld getClassWorld()
    {
        return classWorld;
    }

    public String getWorkingDirectory()
    {
        return workingDirectory;
    }

    public File getMultiModuleProjectDirectory()
    {
        return multiModuleProjectDirectory;
    }

    public boolean isDebug()
    {
        return debug;
    }

    public boolean isQuiet()
    {
        return quiet;
    }

    public boolean isShowErrors()
    {
        return showErrors;
    }

    public Properties getUserProperties()
    {
        return userProperties;
    }

    public Properties getSystemProperties()
    {
        return systemProperties;
    }

    public MavenExecutionRequest getRequest()
    {
        return request;
    }

    public void setUserProperties( Properties properties ) 
    {
        this.userProperties.putAll( properties );      
    }
}

可以看出,这些参数非常丰富,也许可以满足我们的需求,但是CliRequest只有一个默认修饰符的构造方法,也就说只有位于org.apache.maven.cli包下的类才有访问CliRequest构造方法的权限,我们可以在平台项目里新建一个包org.apache.maven.cli,然后再创建一个类(如:DefaultCliRequest)继承自CliRequest,然后实现一个public的构造方法,就可以在任何包里使用该类了,如下代码:

package org.apache.maven.cli;

import org.codehaus.plexus.classworlds.ClassWorld;

public class DefaultCliRequest extends CliRequest{

	public DefaultCliRequest(String[] args, ClassWorld classWorld) {
		super(args, classWorld);
	}
	
	public void setWorkingDirectory(String directory) {
		this.workingDirectory = directory;
	}
}

定义好参数类型DefaultCliRequest后,我们再来看看打包的代码:

public void doPackage() {
	String[] commands = new String[] { "clean", "package", "-Dmaven.test.skip" };
	DefaultCliRequest request = new DefaultCliRequest(commands, null);
	request.setWorkingDirectory("D:/hello/pom.xml");

	Repository repository = new Repository();
	repository.setId("nexus");
	repository.setName("nexus");
	repository.setUrl("https://repo.maven.apache.org/maven2");
	RepositoryPolicy policy = new RepositoryPolicy();
	policy.setEnabled(true);
	policy.setUpdatePolicy("always");
	policy.setChecksumPolicy("fail");
	repository.setReleases(policy);
	repository.setSnapshots(policy);

	String javaVesion = "11";
	Profile profile = new Profile();
	profile.setId("java11");
	Activation activation = new Activation();
	activation.setActiveByDefault(true);
	activation.setJdk(javaVesion);
	profile.setActivation(activation);
	profile.setRepositories(Arrays.asList(repository));
	profile.setPluginRepositories(Arrays.asList(repository));

	Properties properties = new Properties();
	properties.put("java.home", "D:/java/jdk-11.0.16.2");
	properties.put("java.version", javaVesion);
	properties.put("maven.compiler.source", javaVesion);
	properties.put("maven.compiler.target", javaVesion);
	properties.put("maven.compiler.compilerVersion", javaVesion);
	properties.put("project.build.sourceEncoding", "UTF-8");
	properties.put("project.reporting.outputEncoding", "UTF-8");
	profile.setProperties(properties);
	MavenExecutionRequest executionRequest = request.getRequest();
	executionRequest.setProfiles(Arrays.asList(profile));

	MavenCli cli = new MavenCli();
	try {
		cli.doMain(request);
	} catch (Exception e) {
		e.printStackTrace();
	}
}

如果需要设置其他参数,也可以通过以上参数自行添加。

一般情况下,我们在Docker环境中通过Docker命令来制作镜像,过程如下:
1.首先编写Dockerfile文件;
2.通过docker build制作镜像;
3.通过docker push上传镜像;
可以看出,如果要使用docker制作镜像的话,必须要有docker环境,而且需要编写Dockerfile文件。当然,也可以不用安装docker环境,直接使用doker的远程接口:post/build。但是,在远程服务器中仍然需要安装doker环境和编写Dockerfile。在不依赖Docker环境的情况下,仍然可以制作镜像,下面就介绍一款工具Jib的用法。
Jib是谷歌开源的一套工具,github地址,它是一个无需Docker守护进程——也无需深入掌握Docker最佳实践的情况下,为Java应用程序构建Docker和OCI镜像, 它可以作为Maven和Gradle的插件,也可以作为Java库。

比如,使用jib-maven-plugin插件构建镜像的代码如下:

<plugin>
	<groupId>com.google.cloud.tools</groupId>
	<artifactId>jib-maven-plugin</artifactId>
	<version>3.3.0</version>
	<configuration>
		<from>
			<image>openjdk:13-jdk-alpine</image>
		</from>
		<to>
			<image>gcr.io/dhorse/client</image>
			<tags>
				<tag>102</tag>
			</tags>
			<auth>
				<!--连接镜像仓库的账号和密码 -->
				<username>username</username>
				<password>password</password>
			</auth>
		</to>
		<container>
			<ports>
				<port>8080</port>
			</ports>
		</container>
	</configuration>
	<executions>
		<execution>
			<phase>package</phase>
			<goals>
				<goal>build</goal>
			</goals>
		</execution>
	</executions>
</plugin>

然后使用命令进行构建:

mvn compile jib:build

可以看出,无需docker环境就可以实现镜像的构建。但是,要想通过平台类型的系统去为每个系统构建镜像,显然通过插件的方式,不太合适,因为需要每个被构建系统引入jib-maven-plugin插件才行,也就是需要改造每一个系统,这样就会带来一定的麻烦。那么有没有不需要改造系统的方式直接进行构建镜像呢?答案是通过Jib-core就可以实现。

首先,在使用Jib-core的项目中引入依赖,maven如下:

<dependency>
	<groupId>com.google.cloud.tools</groupId>
	<artifactId>jib-core</artifactId>
	<version>0.22.0</version>
</dependency>

然后就可以直接使用Jib-core的API来进行制作镜像,如下代码:

try {
	JibContainerBuilder jibContainerBuilder = null;
	if (StringUtils.isBlank(context.getProject().getBaseImage())) {
		jibContainerBuilder = Jib.fromScratch();
	} else {
		jibContainerBuilder = Jib.from(context.getProject().getBaseImage());
	}
	//连接镜像仓库5秒超时
	System.setProperty("jib.httpTimeout", "5000");
	System.setProperty("sendCredentialsOverHttp", "true");
	String fileNameWithExtension = targetFiles.get(0).toFile().getName();
	List<String> entrypoint = Arrays.asList("java", "-jar", fileNameWithExtension);
	RegistryImage registryImage = RegistryImage.named(context.getFullNameOfImage()).addCredential(
			context.getGlobalConfigAgg().getImageRepo().getAuthUser(),
			context.getGlobalConfigAgg().getImageRepo().getAuthPassword());
	jibContainerBuilder.addLayer(targetFiles, "/")
		.setEntrypoint(entrypoint)
		.addVolume(AbsoluteUnixPath.fromPath(Paths.get("/etc/localtime")))
		.containerize(Containerizer.to(registryImage)
				.setAllowInsecureRegistries(true)
				.addEventHandler(LogEvent.class, logEvent -> logger.info(logEvent.getMessage())));
} catch (Exception e) {
	logger.error("Failed to build image", e);
	return false;
}

其中,targetFiles是要构建镜像的目标文件,比如springboot打包后的jar文件。
通过Jib-core,可以很轻松的实现镜像构建,而不需要依赖任何其他环境,也不需要被构建系统做任何改造,非常方便。

类似代码仓库提供的Restful API,也可以通过Restful API来操作镜像仓库,以Harbor创建一个项目为例,代码如下:

public void createProject(ImageRepo imageRepo) {
	String uri = "api/v2.0/projects";
	if(!imageRepo.getUrl().endsWith("/")) {
		uri = "/" + uri;
	}
	HttpPost httpPost = new HttpPost(imageRepo.getUrl() + uri);
	RequestConfig requestConfig = RequestConfig.custom()
			.setConnectionRequestTimeout(5000)
			.setConnectTimeout(5000)
			.setSocketTimeout(5000)
			.build();
	httpPost.setConfig(requestConfig);
	httpPost.setHeader("Content-Type", "application/json;charset=UTF-8");
	httpPost.setHeader("Authorization", "Basic "+ Base64.getUrlEncoder().encodeToString((imageRepo.getAuthUser() + ":" + imageRepo.getAuthPassword()).getBytes()));
	ObjectNode objectNode = JsonUtils.getObjectMapper().createObjectNode();
	objectNode.put("project_name", "dhorse");
	//1:公有类型
	objectNode.put("public", 1);
	httpPost.setEntity(new StringEntity(objectNode.toString(),"UTF-8"));
	try (CloseableHttpResponse response = createHttpClient(imageRepo.getUrl()).execute(httpPost)){
		if (response.getStatusLine().getStatusCode() != 201
				&& response.getStatusLine().getStatusCode() != 409) {
			LogUtils.throwException(logger, response.getStatusLine().getReasonPhrase(),
					MessageCodeEnum.IMAGE_REPO_PROJECT_FAILURE);
		}
	} catch (IOException e) {
		LogUtils.throwException(logger, e, MessageCodeEnum.IMAGE_REPO_PROJECT_FAILURE);
	}
}

k8s集群

同样,k8s也提供了Restful API。同时,官方也提供了各种语言的客户端,下面以Java语言的客户端为例,来创建一个deployment。
首先,引入Maven依赖:

<dependency>
	<groupId>io.kubernetes</groupId>
	<artifactId>client-java</artifactId>
	<version>13.0.0</version>
</dependency>

然后,使用如下代码:

public boolean createDeployment(DeployContext context) {
	V1Deployment deployment = new V1Deployment();
	deployment.apiVersion("apps/v1");
	deployment.setKind("Deployment");
	deployment.setMetadata(deploymentMetaData(context.getDeploymentAppName()));
	deployment.setSpec(deploymentSpec(context));
	ApiClient apiClient = this.apiClient(context.getCluster().getClusterUrl(),
			context.getCluster().getAuthToken(), 1000, 1000);
	AppsV1Api api = new AppsV1Api(apiClient);
	CoreV1Api coreApi = new CoreV1Api(apiClient);
	String namespace = context.getProjectEnv().getNamespaceName();
	String labelSelector = K8sUtils.getDeploymentLabelSelector(context.getDeploymentAppName());
	try {
		V1DeploymentList oldDeployment = api.listNamespacedDeployment(namespace, null, null, null, null,
				labelSelector, null, null, null, null, null);
		if (CollectionUtils.isEmpty(oldDeployment.getItems())) {
			deployment = api.createNamespacedDeployment(namespace, deployment, null, null, null);
		} else {
			deployment = api.replaceNamespacedDeployment(context.getDeploymentAppName(), namespace, deployment, null, null,
					null);
		}
	} catch (ApiException e) {
		if (!StringUtils.isBlank(e.getMessage())) {
			logger.error("Failed to create k8s deployment, message: {}", e.getMessage());
		} else {
			logger.error("Failed to create k8s deployment, message: {}", e.getResponseBody());
		}
		return false;
	}
	return true;
}

private ApiClient apiClient(String basePath, String accessToken, int connectTimeout, int readTimeout) {
	ApiClient apiClient = new ClientBuilder().setBasePath(basePath).setVerifyingSsl(false)
			.setAuthentication(new AccessTokenAuthentication(accessToken)).build();
	apiClient.setConnectTimeout(connectTimeout);
	apiClient.setReadTimeout(readTimeout);
	return apiClient;
}

至此,关键的技术点已经介绍完了,更多内容,请参考这里


About Joyk


Aggregate valuable and interesting links.
Joyk means Joy of geeK