jclasslib修改字节码并重打jar包

2021年9月15日 0 条评论 1.24k 次阅读 0 人点赞

背景

北京IDC托管价格越来越贵，公司决定将机房迁移到环京地区。最近在参与公司机房迁移工作，主要负责实时计算和用户画像两个方面。用户画像任务还是很多的，在收尾阶段发现一个隐藏很深的任务，git上已经找不到源代码了，正常运行的任务只有一个jar包。尝试将jar包直接提交到新集群上，任务运行十分钟后抛出Exception，有可能代码里把HDFS的Nameservice直接写死了，需要反编译进行确认。

Shell
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
13-09-2021 09:56:03 CST dumpMongo INFO - Exception in thread "main" java.lang.IllegalArgumentException: java.net.UnknownHostException: pink
13-09-2021 09:56:03 CST dumpMongo INFO - at org.apache.hadoop.security.SecurityUtil.buildTokenService(SecurityUtil.java:374)
13-09-2021 09:56:03 CST dumpMongo INFO - at org.apache.hadoop.hdfs.NameNodeProxies.createNonHAProxy(NameNodeProxies.java:310)
13-09-2021 09:56:03 CST dumpMongo INFO - at org.apache.hadoop.hdfs.NameNodeProxies.createProxy(NameNodeProxies.java:176)
13-09-2021 09:56:03 CST dumpMongo INFO - at org.apache.hadoop.hdfs.DFSClient.<init>(DFSClient.java:707)
13-09-2021 09:56:03 CST dumpMongo INFO - at org.apache.hadoop.hdfs.DFSClient.<init>(DFSClient.java:650)
13-09-2021 09:56:03 CST dumpMongo INFO - at org.apache.hadoop.hdfs.DistributedFileSystem.initialize(DistributedFileSystem.java:148)
13-09-2021 09:56:03 CST dumpMongo INFO - at org.apache.hadoop.fs.FileSystem.createFileSystem(FileSystem.java:2643)
13-09-2021 09:56:03 CST dumpMongo INFO - at org.apache.hadoop.fs.FileSystem.access$200(FileSystem.java:93)
13-09-2021 09:56:03 CST dumpMongo INFO - at org.apache.hadoop.fs.FileSystem$Cache.getInternal(FileSystem.java:2680)
13-09-2021 09:56:03 CST dumpMongo INFO - at org.apache.hadoop.fs.FileSystem$Cache.get(FileSystem.java:2662)
13-09-2021 09:56:03 CST dumpMongo INFO - at org.apache.hadoop.fs.FileSystem.get(FileSystem.java:379)
13-09-2021 09:56:03 CST dumpMongo INFO - at DumpMongoJob$.main(DumpMongoJob.scala:64)
13-09-2021 09:56:03 CST dumpMongo INFO - at DumpMongoJob.main(DumpMongoJob.scala)
13-09-2021 09:56:03 CST dumpMongo INFO - at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
13-09-2021 09:56:03 CST dumpMongo INFO - at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
13-09-2021 09:56:03 CST dumpMongo INFO - at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
13-09-2021 09:56:03 CST dumpMongo INFO - at java.lang.reflect.Method.invoke(Method.java:498)
13-09-2021 09:56:03 CST dumpMongo INFO - at org.apache.spark.deploy.JavaMainApplication.start(SparkApplication.scala:52)
13-09-2021 09:56:03 CST dumpMongo INFO - at org.apache.spark.deploy.SparkSubmit.org$apache$spark$deploy$SparkSubmit$$runMain(SparkSubmit.scala:845)
13-09-2021 09:56:03 CST dumpMongo INFO - at org.apache.spark.deploy.SparkSubmit.doRunMain$1(SparkSubmit.scala:161)
13-09-2021 09:56:03 CST dumpMongo INFO - at org.apache.spark.deploy.SparkSubmit.submit(SparkSubmit.scala:184)
13-09-2021 09:56:03 CST dumpMongo INFO - at org.apache.spark.deploy.SparkSubmit.doSubmit(SparkSubmit.scala:86)
13-09-2021 09:56:03 CST dumpMongo INFO - at org.apache.spark.deploy.SparkSubmit$$anon$2.doSubmit(SparkSubmit.scala:920)
13-09-2021 09:56:03 CST dumpMongo INFO - at org.apache.spark.deploy.SparkSubmit$.main(SparkSubmit.scala:929)
13-09-2021 09:56:03 CST dumpMongo INFO - at org.apache.spark.deploy.SparkSubmit.main(SparkSubmit.scala)
13-09-2021 09:56:03 CST dumpMongo INFO - Caused by: java.net.UnknownHostException: pink
13-09-2021 09:56:03 CST dumpMongo INFO - ... 25 more
13-09-2021 09:56:03 CST dumpMongo INFO - 21/09/13 09:56:03 INFO spark.SparkContext: Invoking stop() from shutdown hook
13-09-2021 09:56:03 CST dumpMongo INFO - 21/09/13 09:56:03 INFO server.AbstractConnector: Stopped Spark@30e92cb9{HTTP/1.1,[http/1.1]}{0.0.0.0:4047}

Java class文件反编译

支持“.class”文件反编译的工具有很多，比如IDEA、JD-GUI、Recaf等。在解决上述问题时本人首选IDEA，具体步骤如下：

  • 打开任意一个工程（建议新建），在src同级目录下建立“lib”文件夹并移动jar包到该目录

  • 右键点击lib文件夹并在弹出的选项中选择“Add as Library”

添加完成后可以看到jar已经可以支持浏览了，找到异常中提示的DumpMongoJob（Scala）并反编译成java代码：

从反编译后的文件中可以看到，HDFS的Nameservice已经写死为固定值Pink:

Java
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
SparkConf sparkConf = (new SparkConf()).setAppName(this.getClass().getSimpleName());
SparkSession spark = .MODULE$.builder().enableHiveSupport().config(sparkConf).getOrCreate();
OptionParser optionParser = new OptionParser();
optionParser.parse(args);
Config conf = optionParser.getConf();
String mongoDbUri = (String)conf.getOptString("mongo.db.uri").get();
String execSql = (String)conf.getOptString("dim.exec.sql").get();
String output = (String)conf.getOptString("output.path").get();
FileSystem hdfs = FileSystem.get(new URI("hdfs://pink"), new Configuration());
Path p = new Path(output);
if (hdfs.exists(p) && output.contains("/user/somebody/other_path")) {
   BoxesRunTime.boxToBoolean(hdfs.delete(p, true));
} else {
   BoxedUnit var10000 = BoxedUnit.UNIT;
}

jclasslib修改字节码

定为具体原因后就需要对jar包中对应的.class文件进行编辑并二次打包，将jar包进行解压并将DumpMongoJob$.class加载到jclasslib bytecode viewer中：

通过查找可以定位到包含“pink”常量的字节码位置，点击编辑将“pink”改为目标集群名称最后保存即可。

编辑后替换原来的.class文件并再次打包，此时提交到集群上后发现运行正常，可以正常将数据落到HDFS上。

Shell
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
15-09-2021 20:02:23 CST dumpMongo INFO - 21/09/15 20:02:23 INFO cluster.YarnScheduler: Removed TaskSet 1.0, whose tasks have all completed, from pool
15-09-2021 20:02:23 CST dumpMongo INFO - 21/09/15 20:02:23 INFO scheduler.DAGScheduler: ResultStage 1 (runJob at SparkHadoopWriter.scala:78) finished in 159.193 s
15-09-2021 20:02:23 CST dumpMongo INFO - 21/09/15 20:02:23 INFO scheduler.DAGScheduler: Job 1 finished: runJob at SparkHadoopWriter.scala:78, took 159.205425 s
15-09-2021 20:02:23 CST dumpMongo INFO - 21/09/15 20:02:23 INFO io.SparkHadoopWriter: Job job_20210922195858_0015 committed.
15-09-2021 20:02:23 CST dumpMongo INFO - 21/09/15 20:02:23 INFO server.AbstractConnector: Stopped Spark@790132f7{HTTP/1.1,[http/1.1]}{0.0.0.0:4042}
15-09-2021 20:02:23 CST dumpMongo INFO - 21/09/15 20:02:23 INFO ui.SparkUI: Stopped Spark web UI at http://10.10.41.1:4042
15-09-2021 20:02:23 CST dumpMongo INFO - 21/09/15 20:02:23 INFO cluster.YarnClientSchedulerBackend: Interrupting monitor thread
15-09-2021 20:02:23 CST dumpMongo INFO - 21/09/15 20:02:23 INFO cluster.YarnClientSchedulerBackend: Shutting down all executors
15-09-2021 20:02:23 CST dumpMongo INFO - 21/09/15 20:02:23 INFO cluster.YarnSchedulerBackend$YarnDriverEndpoint: Asking each executor to shut down
15-09-2021 20:02:23 CST dumpMongo INFO - 21/09/15 20:02:23 INFO cluster.SchedulerExtensionServices: Stopping SchedulerExtensionServices
15-09-2021 20:02:23 CST dumpMongo INFO - (serviceOption=None,
15-09-2021 20:02:23 CST dumpMongo INFO -  services=List(),
15-09-2021 20:02:23 CST dumpMongo INFO -  started=false)
15-09-2021 20:02:23 CST dumpMongo INFO - 21/09/15 20:02:23 INFO cluster.YarnClientSchedulerBackend: Stopped
15-09-2021 20:02:23 CST dumpMongo INFO - 21/09/15 20:02:23 INFO spark.MapOutputTrackerMasterEndpoint: MapOutputTrackerMasterEndpoint stopped!
15-09-2021 20:02:23 CST dumpMongo INFO - 21/09/15 20:02:23 INFO memory.MemoryStore: MemoryStore cleared
15-09-2021 20:02:23 CST dumpMongo INFO - 21/09/15 20:02:23 INFO storage.BlockManager: BlockManager stopped
15-09-2021 20:02:23 CST dumpMongo INFO - 21/09/15 20:02:23 INFO storage.BlockManagerMaster: BlockManagerMaster stopped
15-09-2021 20:02:23 CST dumpMongo INFO - 21/09/15 20:02:23 INFO scheduler.OutputCommitCoordinator$OutputCommitCoordinatorEndpoint: OutputCommitCoordinator stopped!
15-09-2021 20:02:23 CST dumpMongo INFO - 21/09/15 20:02:23 INFO spark.SparkContext: Successfully stopped SparkContext
15-09-2021 20:02:23 CST dumpMongo INFO - 21/09/15 20:02:23 INFO util.ShutdownHookManager: Shutdown hook called
15-09-2021 20:02:23 CST dumpMongo INFO - 21/09/15 20:02:23 INFO util.ShutdownHookManager: Deleting directory /tmp/spark-ac0222b4-68a1-4a7d-8ec8-e05c8aee1907
15-09-2021 20:02:23 CST dumpMongo INFO - 21/09/15 20:02:23 INFO util.ShutdownHookManager: Deleting directory /tmp/spark-d6e79675-cad5-400b-985e-72842ed7eb48
15-09-2021 20:02:24 CST dumpMongo INFO - Process completed successfully in 275 seconds.
15-09-2021 20:02:24 CST dumpMongo INFO - Finishing job dumpMongo attempt: 0 at 1632312144837 with status SUCCEEDED

此外ingokegel还提供IDEA 插件版本的jclasslib bytecode viewer，插件版本的好处是不需要解压jar包和二次打包，编辑完成后直接点击保存图标即可。具体使用步骤：

  • 选中需要编辑的“.class”文件；
  • 菜单栏中选中“View”选项卡；
  • 选择“Show Bytecode With jclasslib”；

编辑完成后记得保存再退出。

参考资料

https://github.com/ingokegel/jclasslib

https://plugins.jetbrains.com/plugin/9248-jclasslib-bytecode-viewer

© 著作权归作者所有

勇敢，和生活的艰难无关。

文章评论（0）