【Bigtop】利用Bigtop3.2.0编译大数据组件RPM包

利用Bigtop3.2.0编译大数据组件RPM包

前言

原文参考：Bigtop 从0开始

参考了上述的博文自己尝试了编译组件，过程还是遇到很多问题，一一记录，方便后人。

Bigtop项目官网：BigTop

正文

Mvn本地目录的修改

我在编译过程中启动bigtop镜像的时候，把mvn挂载的目录改了，因此要在修改

/usr/local/maven/conf/settings.xml

时填写相关配置信息，这个一定要注意，不然每次编译都重新下载依赖：

<localRepository>/root/.m2/repository/</localRepository>

Flink

这个是目前困扰我时间最久的一个组件，Flink的编译过程中，会去下载nodejs，但是nodejs.org被墙了，所以每次编译到runtime-web都会超时失败：

INFO:I/O exception (java.net.SocketException) caught when processing request to{s}-https://nodejs.org:443:Network is unreachable (connect failed)

解决办法是手动下载版本包放到mvn仓库下，但是这里有个坑，如果直接使用v16.13.2版本的node，会在这一步成功后，卡在下一步，并且报一个权限异常的问题：

/ws/build/flink/rpm/BuLD/flink-1.15.3/flink-runtime-web/web-dashboard/node/npm: Permission denied

因此，整体的解决方法是，修改flink源码包中的

flink-runtime-web/pom.xml

文件的npm部分为下面这样：

<plugin><groupId>com.github.eirslett</groupId><artifactId>frontend-maven-plugin</artifactId><version>1.11.0</version><executions><execution><id>install node and npm</id><goals><goal>install-node-and-npm</goal></goals><configuration><!--这里修改了node的版本--><nodeVersion>v12.22.1</nodeVersion><npmVersion>6.14.12</npmVersion></configuration></execution><execution><id>npm install husky</id><goals><goal>npm</goal></goals><configuration><arguments>install husky --registry=https://registry.npmmirror.com</arguments></configuration></execution><execution><id>npm install</id><goals><goal>npm</goal></goals><configuration><arguments>install --cache-max=0 --no-save --registry=https://registry.npmmirror.com</arguments><environmentVariables><HUSKY_SKIP_INSTALL>true</HUSKY_SKIP_INSTALL></environmentVariables></configuration></execution><execution><id>npm install local</id><goals><goal>npm</goal></goals><configuration><arguments>install --registry=https://registry.npmmirror.com --force</arguments></configuration></execution><execution><id>npmrun ci-check</id><goals><goal>npm</goal></goals><configuration><arguments>run ci-check</arguments></configuration></execution></executions><configuration><workingDirectory>web-dashboard</workingDirectory></configuration></plugin>

最主要的就是修改nodejs的版本，降到12.22.1手动把nodejs包放到mvn本地目录下：

这里注意下存放在本地的包版本号前不加v，手动下载的时候自带的名称里是有v的，另外，所有组件编译碰到nodejs下载失败的都可以参照这个操作

这样就没啥大问题了，接着就是等待编译通过了，其他没遇到什么大问题：

Kafka

先执行一次kafka的编译，让编译流程自动把kafka的源码包下载下来，

bigtop3.2.0

分支对应的kafka版本是

2.8.1

：

./gradlew kafka-clean kafka-pkg -PparentDir=/usr/bigtop -PpkgSuffix-PbuildThreads=16C repo 

然后我们开始做修改。

grgit版本

首先是grgit的版本要修改一下，默认用的4.1.0，对应的pom文件已经404了，对应PR在这里：MINOR: Bump version of grgit to 4.1.1

修改

${SOURCE_CODE}/gradle/dependencies.gradle

中grgit的版本：

versions +=[
  activation:"1.1.1",
  apacheda:"1.0.2",
  apacheds:"2.0.0-M24",
  argparse4j:"0.7.0",
  bcpkix:"1.66",
  checkstyle:"8.36.2",
  commonsCli:"1.4",
  gradle:"6.8.1",
  gradleVersionsPlugin:"0.36.0",
  grgit:"4.1.1",// 修改这一行
  httpclient:"4.5.13",]

手动准备gradle的文件

先手动下载gradle-wrapper.jar到

${SOURCE_CODE}/gradle/wrapper/gradle-wrapper.jar

，这里用的版本是6.8.1

curl-s-S--retry3-L-o"gradle-wrapper.jar"  https://mirror.ghproxy.com/https://raw.githubusercontent.com/gradle/gradle/v6.8.1/gradle/wrapper/gradle-wrapper.jar

然后把

gradle-6.8.1-all.zip

文件准备好，做成个http的服务，我这里直接用python启了一个SimpleHTTPServer，接着我们修改

${SOURCE_CODE}/gradle/wrapper/gradle-wrapper.properties

文件：

distributionBase=GRADLE_USER_HOME
distributionPath=wrapper/dists
# 修改这里为对应的http服务distributionUrl=http://172.18.2.31:444/gradle-6.8.1-all.zip
zipStoreBase=GRADLE_USER_HOME
zipStorePath=wrapper/dists

Hadoop

Bigtop3.2.0使用的是hadoop3.3.4，YarnUI的编译过程会使用到nodejs，版本默认也用的是12.22.1，如果之前编译flink时已经把版本包下载到本地，就不用再做额外操作了，否则参考Flink章节把nodejs先存到本地。

修改mvn仓库

修改

${SOURCE_CODE}/pom.xml

中的mvn配置：

<distMgmtSnapshotsId>apache.snapshots.https</distMgmtSnapshotsId><distMgmtSnapshotsName>Apache Development Snapshot Repository</distMgmtSnapshotsName><!--这一行修改为阿里的--><distMgmtSnapshotsUrl>https://maven.aliyun.com/repository/apache-snapshots</distMgmtSnapshotsUrl><distMgmtStagingId>apache.staging.https</distMgmtStagingId><distMgmtStagingName>Apache Release Distribution Repository</distMgmtStagingName><!--这一行修改为阿里的--><distMgmtStagingUrl>https://maven.aliyun.com/repository/central</distMgmtStagingUrl>

合入Patch

这里要手动合一个Patch，否则编译YarnUI的时候还是会报如下的错：

[INFO] error [email protected]: The engine "node" is incompatible with this module. Expected version ">= 14.0.0". Got "12.22.1"

对应的issue：hadoop-deb FAILED on project hadoop-yarn-applications-catalog-webapp

将一下内容保存到

bigtop-packages/src/common/hadoop/patch8-YARN-11528-triple-beam.diff

：

diff--git a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-applications/hadoop-yarn-applications-catalog/hadoop-yarn-applications-catalog-webapp/package.json b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-applications/hadoop-yarn-applications-catalog/hadoop-yarn-applications-catalog-webapp/package.json
index f09442cfc4e87..59cc3da179fd0 100644
--- a/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-applications/hadoop-yarn-applications-catalog/hadoop-yarn-applications-catalog-webapp/package.json
+++ b/hadoop-yarn-project/hadoop-yarn/hadoop-yarn-applications/hadoop-yarn-applications-catalog/hadoop-yarn-applications-catalog-webapp/package.json
@@ -19,6 +19,9 @@
         "shelljs":"^0.2.6",
         "apidoc":"0.17.7"},
+    "resolutions":{
+        "triple-beam":"1.3.0"
+    },
     "scripts":{"prestart":"npm install & mvn clean package",
         "pretest":"npm install"

然后执行编译即可：

Tez

phantomjs下载不了

过程中会从github拉取phantomjs的安装压缩包，大概率会超时失败：

Saving to /tmp/phantomjs/phantomjs-2.1.1-linux-x86_64.tar.bz2\nReceiving...\n\nError making request.
Error: socket hang up
    at TLSSocket.onHangUp (_tls_wrap.js:1097:19)
    at TLSSocket.g (events.js:273:16)
    at emitNone (events.js:85:20)
    at TLSSocket.emit (events.js:179:7)
    at endReadableNT (_stream_readable.js:913:12)
    at _combinedTickCallback (internal/process/next_tick.js:74:11)
    at process._tickCallback (internal/process/next_tick.js:98:9)
    
Please report this full log at https://github.com/Medium/phantomjs"

网上各种让用npm手动安装的，这里其实只需要手动下载一下放在

/tmp/phantomjs/phantomjs-2.1.1-linux-x86_64.tar.bz2

即可，编译过程自己就会拿tmp目录下的压缩包

allow-root

默认pom.xml把sudo禁用了，会得到下面的报错：

[INFO] Running 'bower install --allow-root=false'in /ws/build/tez/rpm/BUILD/apache-tez-0.10.1-src/tez-ui/src/main/webapp
[ERROR] bower ESUDO         Cannot be run with sudo[ERROR][ERROR] Additional error details:
[ERROR] Since bower is a user command, there is no need to execute it with superuser permissions.
[ERROR] If you're having permission errors when using bower without sudo, please spend a few minutes learning more about how your system should work and make any necessary repairs.
[ERROR][ERROR] http://www.joyent.com/blog/installing-node-and-npm
[ERROR] https://gist.github.com/isaacs/579814
[ERROR][ERROR] You can however run a command with sudo using "--allow-root" option

手动把allow-root打开，改为true

然后一把过：

Zeppelin

替换下载路径

变异过程中zeppelin会去下载部分组件的源码包，直接连的apache源，很慢：

这里最好替换成国内的原，但是版本比较老，阿里和清华的源已经没了，这里目前尝试可以替换成华为源：

rlang/pom.xml和spark/pom.xml

<spark.src.download.url>
            https://mirrors.huaweicloud.com/apache/spark/${spark.archive}/${spark.archive}.tgz
        </spark.src.download.url><spark.bin.download.url>
            https://mirrors.huaweicloud.com/apache/spark/${spark.archive}/${spark.archive}-bin-without-hadoop.tgz
        </spark.bin.download.url>

flink/flink-scala-2.11/flink-scala-parent/pom.xml
flink/flink-scala-parent/pom.xml
flink/flink-scala-2.12/flink-scala-parent/pom.xml

<properties><flink.bin.download.url>https://mirrors.huaweicloud.com/apache/flink/flink-${flink.version}/flink-${flink.version}-bin-scala_${flink.scala.binary.version}.tgz</flink.bin.download.url></properties>

git设置

过程中有个前端依赖会从git拉取代码：

这样拉不到，要配下git的配置：

git config --global url.https://gh-proxy.com/https://github.com/.insteadOf git://github.com/

OK，接下来等编译通过：