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。

image.png

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：

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

Dload Upload Total Spent Left Speed

100 34 0 34 0 0 2314 0 --:--:-- --:--:-- --:--:-- 2428

{

"from": 0,

"sessions": [],

"total": 0

}

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

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

{"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：

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

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

Dload Upload Total Spent Left Speed

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

{

"id": 0,

"log": [

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

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

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

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

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

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

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

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

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

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

],

"state": "success"

}

此外，还可以通过下面的api，获取日志信息：

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

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

Dload Upload Total Spent Left Speed

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

{

"from": 0,

"id": 3,

"log": [

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

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

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

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

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

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

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

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

"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)",

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

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

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

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

....

....

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

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

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

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

"\t client token: N/A",

"\t diagnostics: N/A",

"\t ApplicationMaster host: 192.168.1.30",

"\t ApplicationMaster RPC port: 0",

"\t queue: root.romain",

"\t start time: 1445416649481",

"\t final status: SUCCEEDED",

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

"\t user: romain",

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

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

],

"total": 67

}

还可以在命令行中添加参数，例如这里计算一百次：

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

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

如果想终止任务，可以调用以下API：

curl -X DELETE localhost:8998/batches/1

{"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：

curl -X DELETE localhost:8998/batches/1

session not found

提交Python任务

提交Python任务和Jar包类似：

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

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

检查任务状态：

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

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

Dload Upload Total Spent Left Speed

100 616 0 616 0 0 77552 0 --:--:-- --:--:-- --:--:-- 88000

{

"id": 2,

"log": [

"\t ApplicationMaster host: 192.168.1.30",

"\t ApplicationMaster RPC port: 0",

"\t queue: root.romain",

"\t start time: 1445417899564",

"\t final status: UNDEFINED",

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

"\t user: romain",

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

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

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

],

"state": "running"

}

获取日志信息：

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