大数据学习（五）：如何使用 Livy提交spark批量任务--转载

Livy是一个开源的REST 接口，用于与Spark进行交互，它同时支持提交执行代码段和完整的程序。

 

Livy封装了spark-submit并支持远端执行。

启动服务器

执行以下命令，启动livy服务器。

./bin/livy-server

这里假设spark使用yarn模式，所以所有文件路径都默认位于HDFS中。如果是本地开发模式的话，直接使用本地文件即可（注意必须配置livy.conf文件，设置livy.file.local-dir-whitelist = directory，以允许文件添加到session）。

提交jar包

首先我们列出当前正在执行的任务：

1. curl localhost:8998/sessions | python -m json.tool % Total % Received % Xferd Average Speed Time Time Time Current

2. Dload Upload Total Spent Left Speed

3. 100 34 0 34 0 0 2314 0 --:--:-- --:--:-- --:--:-- 2428

4. {

5. "from": 0,

6. "sessions": [],

7. "total": 0

8. }

然后提交jar包，假设提交的jar包位于hdfs中，路径为/usr/lib/spark/lib/spark-examples.jar

1. curl -X POST --data '{"file": "/user/romain/spark-examples.jar", "className": "org.apache.spark.examples.SparkPi"}' -H "Content-Type: application/json" localhost:8998/batches

2. {"id":0,"state":"running","log":[]}

返回结果中包括了提交的ID，这里为0，我们可以通过下面的命令查看任务状态：

1. curl localhost:8998/batches/0 | python -m json.tool

2. % Total % Received % Xferd Average Speed Time Time Time Current

3. Dload Upload Total Spent Left Speed

4. 100 902 0 902 0 0 91120 0 --:--:-- --:--:-- --:--:-- 97k

5. {

6. "id": 0,

7. "log": [

8. "15/10/20 16:32:21 INFO ui.SparkUI: Stopped Spark web UI at http://192.168.1.30:4040",

9. "15/10/20 16:32:21 INFO scheduler.DAGScheduler: Stopping DAGScheduler",

10. "15/10/20 16:32:21 INFO spark.MapOutputTrackerMasterEndpoint: MapOutputTrackerMasterEndpoint stopped!",

11. "15/10/20 16:32:21 INFO storage.MemoryStore: MemoryStore cleared",

12. "15/10/20 16:32:21 INFO storage.BlockManager: BlockManager stopped",

13. "15/10/20 16:32:21 INFO storage.BlockManagerMaster: BlockManagerMaster stopped",

14. "15/10/20 16:32:21 INFO scheduler.OutputCommitCoordinator$OutputCommitCoordinatorEndpoint: OutputCommitCoordinator stopped!",

15. "15/10/20 16:32:21 INFO spark.SparkContext: Successfully stopped SparkContext",

16. "15/10/20 16:32:21 INFO util.ShutdownHookManager: Shutdown hook called",

17. "15/10/20 16:32:21 INFO util.ShutdownHookManager: Deleting directory /tmp/spark-6e362908-465a-4c67-baa1-3dcf2d91449c"

18. ],

19. "state": "success"

20. }

此外，还可以通过下面的api，获取日志信息：

1. curl localhost:8998/batches/0/log | python -m json.tool

2. % Total % Received % Xferd Average Speed Time Time Time Current

3. Dload Upload Total Spent Left Speed

4. 100 5378 0 5378 0 0 570k 0 --:--:-- --:--:-- --:--:-- 583k

5. {

6. "from": 0,

7. "id": 3,

8. "log": [

9. "SLF4J: Class path contains multiple SLF4J bindings.",

10. "SLF4J: Found binding in [jar:file:/usr/lib/zookeeper/lib/slf4j-log4j12-1.7.5.jar!/org/slf4j/impl/StaticLoggerBinder.class]",

11. "SLF4J: Found binding in [jar:file:/usr/lib/flume-ng/lib/slf4j-log4j12-1.7.5.jar!/org/slf4j/impl/StaticLoggerBinder.class]",

12. "SLF4J: See http://www.slf4j.org/codes.html#multiple_bindings for an explanation.",

13. "SLF4J: Actual binding is of type [org.slf4j.impl.Log4jLoggerFactory]",

14. "15/10/21 01:37:27 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable",

15. "15/10/21 01:37:27 INFO client.RMProxy: Connecting to ResourceManager at /0.0.0.0:8032",

16. "15/10/21 01:37:27 INFO yarn.Client: Requesting a new application from cluster with 1 NodeManagers",

17. "15/10/21 01:37:27 INFO yarn.Client: Verifying our application has not requested more than the maximum memory capability of the cluster (8192 MB per container)",

18. "15/10/21 01:37:27 INFO yarn.Client: Will allocate AM container, with 1408 MB memory including 384 MB overhead",

19. "15/10/21 01:37:27 INFO yarn.Client: Setting up container launch context for our AM",

20. "15/10/21 01:37:27 INFO yarn.Client: Setting up the launch environment for our AM container",

21. "15/10/21 01:37:27 INFO yarn.Client: Preparing resources for our AM container",

22. ....

23. ....

24. "15/10/21 01:37:40 INFO yarn.Client: Application report for application_1444917524249_0004 (state: RUNNING)",

25. "15/10/21 01:37:41 INFO yarn.Client: Application report for application_1444917524249_0004 (state: RUNNING)",

26. "15/10/21 01:37:42 INFO yarn.Client: Application report for application_1444917524249_0004 (state: FINISHED)",

27. "15/10/21 01:37:42 INFO yarn.Client: ",

28. "\t client token: N/A",

29. "\t diagnostics: N/A",

30. "\t ApplicationMaster host: 192.168.1.30",

31. "\t ApplicationMaster RPC port: 0",

32. "\t queue: root.romain",

33. "\t start time: 1445416649481",

34. "\t final status: SUCCEEDED",

35. "\t tracking URL: http://unreal:8088/proxy/application_1444917524249_0004/A",

36. "\t user: romain",

37. "15/10/21 01:37:42 INFO util.ShutdownHookManager: Shutdown hook called",

38. "15/10/21 01:37:42 INFO util.ShutdownHookManager: Deleting directory /tmp/spark-26cdc4d9-071e-4420-a2f9-308a61af592c"

39. ],

40. "total": 67

41. }

还可以在命令行中添加参数，例如这里计算一百次：

1. curl -X POST --data '{"file": "/usr/lib/spark/lib/spark-examples.jar", "className": "org.apache.spark.examples.SparkPi", "args": ["100"]}' -H "Content-Type: application/json" localhost:8998/batches

2. {"id":1,"state":"running","log":[]}

如果想终止任务，可以调用以下API：

1. curl -X DELETE localhost:8998/batches/1

2. {"msg":"deleted"}

当重复调用上述接口时，什么也不会做，因为任务已经删除了：

1. curl -X DELETE localhost:8998/batches/1

2. session not found

提交Python任务

提交Python任务和Jar包类似：

1. curl -X POST --data '{"file": "/user/romain/pi.py"}' -H "Content-Type: application/json" localhost:8998/batches

2. {"id":2,"state":"starting","log":[]}

检查任务状态：

1. curl localhost:8998/batches/2 | python -m json.tool

2. % Total % Received % Xferd Average Speed Time Time Time Current

3. Dload Upload Total Spent Left Speed

4. 100 616 0 616 0 0 77552 0 --:--:-- --:--:-- --:--:-- 88000

5. {

6. "id": 2,

7. "log": [

8. "\t ApplicationMaster host: 192.168.1.30",

9. "\t ApplicationMaster RPC port: 0",

10. "\t queue: root.romain",

11. "\t start time: 1445417899564",

12. "\t final status: UNDEFINED",

13. "\t tracking URL: http://unreal:8088/proxy/application_1444917524249_0006/",

14. "\t user: romain",

15. "15/10/21 01:58:26 INFO yarn.Client: Application report for application_1444917524249_0006 (state: RUNNING)",

16. "15/10/21 01:58:27 INFO yarn.Client: Application report for application_1444917524249_0006 (state: RUNNING)",

17. "15/10/21 01:58:28 INFO yarn.Client: Application report for application_1444917524249_0006 (state: RUNNING)"

18. ],

19. "state": "running"

20. }

获取日志信息：

curl localhost:8998/batches/2/log |  python -m json.tool