Using the Spark Job to Access DLI Metadata

Scenario

DLI allows you to develop a program to create Spark jobs for operations related to databases, DLI or OBS tables, and table data. This example demonstrates how to develop a job by writing a Java program, and use a Spark job to create a database and table and insert table data.

This feature is currently in beta testing. If you want to use it, submit a service ticket to apply for permission to access DLI metadata using Spark jobs.

Constraints

• You must create a queue to use Spark 3.1 for metadata access.
• The following cases are not supported:
  • If you create a database with a SQL job, you cannot write a program to create tables in that database.

    For example, the testdb database is created using the SQL editor of DLI. A program package for creating the testTable table in the testdb database does not work after it is submitted to a Spark Jar job.

• The following cases are supported:
  • You can create databases and tables in a SQL job, and read and insert data using SQL statements or a Spark program.
  • You can create databases and tables in a Spark job, and read and insert data using SQL statements or a Spark program.

Environment Preparations

Before developing a Spark job to access DLI metadata, set up a development environment that meets the following requirements.

Development Process

The following figure shows the process for developing a Spark job to access DLI metadata.

Figure 1 Development process

Table 2 Process description

No.	Phase	Software Portal	Description
1	Create a queue for general use.	DLI console	The DLI queue is created for running your job.
2	Configure the OBS file.	OBS console	To create an OBS table, you need to upload the file to the OBS bucket. Configure the path for storing DLI metadata. This folder is used to store DLI metadata in spark.sql.warehouse.dir.
3	Create a Maven project and configure the POM file.	IntelliJ IDEA	Write a program to create a DLI or OBS table by referring to the sample code.
4	Write code.
5	Debug, compile, and pack the code into a Jar package.
6	Upload the Jar package to OBS and DLI.	OBS console DLI console	You can upload the generated Spark Jar package to an OBS directory and DLI program package.
7	Create a Spark JAR job.	DLI console	The Spark Jar job is created and submitted on the DLI console.
8	Check execution result of the job.	DLI console	You can check the job status and run logs.

Step 1: Create a Queue for General Purpose

Create a queue before submitting Spark jobs. In this example, we will create a general-purpose queue named sparktest.

1. Log in to the DLI management console.
2. In the navigation pane on the left, choose Resources > Resource Pool.
3. On the displayed page, click Buy Resource Pool in the upper right corner.
4. On the displayed page, set the parameters.
5. In this example, we will buy the resource pool in the CN East-Shanghai2 region. Table 3 describes the parameters.

   Table 3 Parameters

   Parameter	Description	Example Value
   Region	Select a region where you want to buy the elastic resource pool.	CN East-Shanghai2
   Project	Project uniquely preset by the system for each region	Default
   Name	Name of the elastic resource pool	dli_resource_pool
   Specifications	Specifications of the elastic resource pool	Standard
   CU Range	The maximum and minimum CUs allowed for the elastic resource pool	64-64
   CIDR Block	CIDR block the elastic resource pool belongs to. If you use an enhanced datasource connection, this CIDR block cannot overlap that of the data source. Once set, this CIDR block cannot be changed.	172.16.0.0/19
   Enterprise Project	Select an enterprise project for the elastic resource pool.	default

6. Click Buy.
7. Click Submit.
8. In the elastic resource pool list, locate the pool you just created and click Add Queue in the Operation column.
9. Set the basic parameters listed below.

   Table 4 Basic parameters for adding a queue

   Parameter	Description	Example Value
   Name	Name of the queue to add	dli_queue_01
   Type	Type of the queue To execute SQL jobs, select For SQL. To execute Flink or Spark jobs, select For general purpose.	_
   Engine	SQL queue engine. The options include Spark and Trino.	_
   Enterprise Project	Select an enterprise project.	default

10. Click Next and configure scaling policies for the queue.

    Click Create to add a scaling policy with varying priority, period, minimum CUs, and maximum CUs.

    Figure 2 shows the scaling policy configured in this example.

    Figure 2 Configuring a scaling policy when adding a queue
    

    Table 5 Scaling policy parameters

    Parameter	Description	Example Value
    Priority	Priority of the scaling policy in the current elastic resource pool. A larger value indicates a higher priority. In this example, only one scaling policy is configured, so its priority is set to 1 by default.	1
    Period	The first scaling policy is the default policy, and its Period parameter configuration cannot be deleted or modified. The period for the scaling policy is from 00 to 24.	00–24
    Min CU	Minimum number of CUs allowed by the scaling policy	16
    Max CU	Maximum number of CUs allowed by the scaling policy	64

11. Click OK.

Step 2: Configure the OBS Bucket File

1. To create an OBS table, upload data to the OBS bucket directory.
   Use the following sample data to create the testdata.csv file and upload it to an OBS bucket.

   12,Michael
   27,Andy
   30,Justin

2. Log in to the OBS console. On the Buckets page, click the name of the OBS bucket you created. In this example, the bucket name is dli-test-obs01.
3. On the Objects tab, click Upload Object. In the displayed dialog box, upload the testdata.csv file to the root directory of the OBS bucket.
4. In the root directory of the OBS bucket, click Create Folder to create a folder and name it warehousepath. This folder is used to store DLI metadata in spark.sql.warehouse.dir.

Step 3: Create a Maven Project and Configure the POM Dependency

Step 4: Write Code

Write the DliCatalogTest program to create a database, DLI table, and OBS table.

For the sample code, see Java Example Code.

1. Import the dependency.

   import org.apache.spark.sql.SparkSession;

2. Create a SparkSession instance.

   When you create a SparkSession, you need to specify spark.sql.session.state.builder, spark.sql.catalog.class, and spark.sql.extensions parameters as configured in the following example.

   • Spark 2.3.x

     SparkSession spark = SparkSession
                     .builder()
                     .config("spark.sql.session.state.builder", "org.apache.spark.sql.hive.UQueryHiveACLSessionStateBuilder")
                     .config("spark.sql.catalog.class", "org.apache.spark.sql.hive.UQueryHiveACLExternalCatalog")
                     .config("spark.sql.extensions","org.apache.spark.sql.DliSparkExtension")
                     .appName("java_spark_demo")
                     .getOrCreate();

   • Spark 2.4.x

     SparkSession spark = SparkSession 
                      .builder() 
                      .config("spark.sql.session.state.builder", "org.apache.spark.sql.hive.UQueryHiveACLSessionStateBuilder") 
                      .config("spark.sql.catalog.class", "org.apache.spark.sql.hive.UQueryHiveACLExternalCatalog") 
                      .config("spark.sql.extensions","org.apache.spark.sql.DliSparkExtension") 
                      .config("spark.sql.hive.implementation","org.apache.spark.sql.hive.client.DliHiveClientImpl")
                      .appName("java_spark_demo") 
                      .getOrCreate();

   • Spark 3.1.x

     SparkSession spark = SparkSession
                     .builder()
                     .config("spark.sql.session.state.builder", "org.apache.spark.sql.hive.UQueryHiveACLSessionStateBuilder")
                     .config("spark.sql.catalog.class", "org.apache.spark.sql.hive.UQueryHiveACLExternalCatalog")
                     .config("spark.sql.extensions","org.apache.spark.sql.DliSparkExtension")
                     .appName("java_spark_demo")
                     .getOrCreate();

   • Spark 3.3.x

     SparkSession spark = SparkSession
                .builder()           
                .config("spark.sql.session.state.builder", "org.apache.spark.sql.hive.DliLakeHouseBuilder")           
                .config("spark.sql.catalog.class", "org.apache.spark.sql.hive.DliLakeHouseCatalog")           
                .appName("java_spark_demo")           
                .getOrCreate();   

3. Create a database.
   The following sample code shows how to create a database and named test_sparkapp.

   spark.sql("create database if not exists test_sparkapp").collect();

4. Create a DLI table and insert test data.

   spark.sql("drop table if exists test_sparkapp.dli_testtable").collect();
   spark.sql("create table test_sparkapp.dli_testtable(id INT, name STRING)").collect();
   spark.sql("insert into test_sparkapp.dli_testtable VALUES (123,'jason')").collect();
   spark.sql("insert into test_sparkapp.dli_testtable VALUES (456,'merry')").collect();

5. Create an OBS Table. Replace the OBS path in the following example with the path you set in Step 2: Configure the OBS Bucket File.

   spark.sql("drop table if exists test_sparkapp.dli_testobstable").collect();
   spark.sql("create table test_sparkapp.dli_testobstable(age INT, name STRING) using csv options (path 'obs://dli-test-obs01/testdata.csv')").collect();

6. Disable the spark session.

   spark.stop();

Step 5: Debug, Compile, and Pack the Code into a Jar Package.

1. Double-click Maven in the tool bar on the right, and double-click clean and compile to compile the code.
   After the compilation is successful, double-click package.

   Figure 9 Compiling and packaging
   
   The generated JAR package is stored in the target directory. In this example, SparkJarMetadata-1.0-SNAPSHOT.jar is stored in D:\DLITest\SparkJarMetadata\target.

   Figure 10 Exporting the JAR file
   

Step 6: Upload the JAR Package to OBS and DLI

• Spark 3.3 or later:

  You can only set the Application parameter when creating a Spark job and select the required JAR file from OBS.

  1. Log in to the OBS console and upload the JAR file to the OBS path.
  2. Log in to the DLI console. In the navigation pane, choose Job Management > Spark Jobs.
  3. Locate the row containing a desired job and click Edit in the Operation column.
  4. Set Application to the OBS path in 1.
     Figure 11 Configuring the application
     
• Versions earlier than Spark 3.3:

  Upload the JAR file to OBS and DLI.

  1. Log in to the OBS console and upload the JAR file to the OBS path.
  2. Upload the file to DLI for package management.
     1. Log in to the DLI management console and choose Data Management > Package Management.
     2. On the Package Management page, click Create in the upper right corner.
     3. In the Create Package dialog, set the following parameters:
        1. Type: Select JAR.
        2. OBS Path: Specify the OBS path for storing the package.
        3. Set Group and Group Name as required for package identification and management.
     4. Click OK.
        Figure 12 Creating a package
        

Step 7: Create a Spark Jar Job

1. Log in to the DLI console. In the navigation pane, choose Job Management > Spark Jobs.
2. On the Spark Jobs page, click Create Job in the upper right corner.
3. On the job creation page, set the job parameters.
   Table 6 describes the parameters. Retain the default values for other parameters.

   Table 6 Spark Jar job parameters

   Parameter	Value
   Queue	Select the DLI queue created for general purpose. For example, select the queue sparktest created in Step 1: Create a Queue for General Purpose.
   Spark Version	Select a Spark version. Select a supported Spark version from the drop-down list. The latest version is recommended.
   Job Name (--name)	Name of a custom Spark Jar job. For example, SparkTestMeta.
   Application	Select the package uploaded to DLI in Step 6: Upload the JAR Package to OBS and DLI. For example, select SparkJarMetadata-1.0-SNAPSHOT.jar.
   Main Class (--class)	The format is program package name + class name. For example, com.huawei.dli.demo.DliCatalogTest.
   Spark Arguments (--conf)	spark.dli.metaAccess.enable=true spark.sql.warehouse.dir=obs://dli-test-obs01/warehousepath NOTE: Set spark.sql.warehouse.dir to the OBS path that is specified in Step 2: Configure the OBS Bucket File.
   Access Metadata	Select Yes.

4. Click Execute to submit the Spark Jar job. On the Spark Jobs page, check the status of the job you submitted.
   Figure 13 Checking the job execution status
   

Step 8: Check Job Execution Result

1. On the Spark Jobs page, check the status of the job you submitted. The initial status is Starting.
2. If the job is successfully executed, the job status is Finished. Perform the following operations to check the created database and table:
   1. On the DLI console, choose SQL Editor in the left navigation pane. The created database test_sparkapp is displayed in the database list.
      Figure 14 Checking the created database
      
   2. Double-click the database name to check the created DLI and OBS tables in the database.
      Figure 15 Checking a table
      
   3. Double-click dli_testtable and click Execute to query data in the table.
      Figure 16 Querying data in the table
      
   4. Comment out the statement for querying the DLI table, double-click the OBS table dli_testobstable, and click Execute to query the OBS table data.
      Figure 17 Querying data in the OBS table
      
3. If the job is successfully executed, the job status is Failed. Click More in the Operation column and select Driver Logs to check the run log.
   Figure 18 Checking logs
   

   After the fault is rectified, click Edit in the Operation column of the job, modify job parameters, and click Execute to run the job again.

Follow-up Guide

• If you want to use Spark Jar jobs to access other data sources, see Using Spark Jobs to Access Data Sources of Datasource Connections.
• For details about the syntax for creating a DLI table, see Creating a DLI Table. For details about the syntax for creating an OBS table, see Creating an OBS Table.
• If you submit the job by calling an API, perform the following operations:

  Call the API for creating a batch processing job. The following table describes the request parameters.

  For details about API parameters, see Creating a Batch Processing Job.

  • Set catalog_name in the request to dli.
  • Add "spark.dli.metaAccess.enable":"true" to the CONF file.

    Configure "spark.sql.warehouse.dir": "obs://bucket/warehousepath" in the CONF file if you need to run the DDL.

    The following example provided you with the complete API request.

    {
        "queue":"citest",
        "file":"SparkJarMetadata-1.0-SNAPSHOT.jar",
        "className":"DliCatalogTest",
        "conf":{"spark.sql.warehouse.dir": "obs://bucket/warehousepath",
        "spark.dli.metaAccess.enable":"true"},
        "sc_type":"A",
        "executorCores":1,
        "numExecutors":6,
        "executorMemory":"4G",
        "driverCores":2,
        "driverMemory":"7G",
        "catalog_name": "dli"
    }

Java Example Code

This example uses Java for coding. The complete sample code is as follows:

package com.huawei.dli.demo;

import org.apache.spark.sql.SparkSession;

public class DliCatalogTest {
    public static void main(String[] args) {

        SparkSession spark = SparkSession
                .builder()
                .config("spark.sql.session.state.builder", "org.apache.spark.sql.hive.UQueryHiveACLSessionStateBuilder")
                .config("spark.sql.catalog.class", "org.apache.spark.sql.hive.UQueryHiveACLExternalCatalog")
                .config("spark.sql.extensions","org.apache.spark.sql.DliSparkExtension")
                .appName("java_spark_demo")
                .getOrCreate();

        spark.sql("create database if not exists test_sparkapp").collect();
        spark.sql("drop table if exists test_sparkapp.dli_testtable").collect();
        spark.sql("create table test_sparkapp.dli_testtable(id INT, name STRING)").collect();
        spark.sql("insert into test_sparkapp.dli_testtable VALUES (123,'jason')").collect();
        spark.sql("insert into test_sparkapp.dli_testtable VALUES (456,'merry')").collect();

        spark.sql("drop table if exists test_sparkapp.dli_testobstable").collect();
        spark.sql("create table test_sparkapp.dli_testobstable(age INT, name STRING) using csv options (path 'obs://dli-test-obs01/testdata.csv')").collect();


        spark.stop();

    }
}

Scala Example Code

object DliCatalogTest {
  def main(args:Array[String]): Unit = {
    val sql = args(0)
    val runDdl =
Try(args(1).toBoolean).getOrElse(true)
    System.out.println(s"sql is $sql
runDdl is $runDdl")
    val sparkConf = new SparkConf(true)
    sparkConf    
      .set("spark.sql.session.state.builder","org.apache.spark.sql.hive.UQueryHiveACLSessionStateBuilder")
      .set("spark.sql.catalog.class","org.apache.spark.sql.hive.UQueryHiveACLExternalCatalog") 
    sparkConf.setAppName("dlicatalogtester")

    val spark = SparkSession.builder
      .config(sparkConf)
      .enableHiveSupport()
      .config("spark.sql.extensions","org.apache.spark.sql.DliSparkExtension")
      .appName("SparkTest")
      .getOrCreate()

    System.out.println("catalog is "
+ spark.sessionState.catalog.toString)
    if (runDdl) {
      val df = spark.sql(sql).collect()
    } else {
      spark.sql(sql).show()
    }

    spark.close()
  }

}

Example Python Code

#!/usr/bin/python
# -*- coding: UTF-8 -*-

from __future__ import print_function

import sys

from pyspark.sql import SparkSession

if __name__ == "__main__":
    url = sys.argv[1]
    creatTbl = "CREATE TABLE test_sparkapp.dli_rds USING JDBC OPTIONS ('url'='jdbc:mysql://%s'," \
              "'driver'='com.mysql.jdbc.Driver','dbtable'='test.test'," \
              " 'passwdauth' = 'DatasourceRDSTest_pwd','encryption' = 'true')" % url

    spark = SparkSession \
        .builder \
        .enableHiveSupport() \
.config("spark.sql.session.state.builder","org.apache.spark.sql.hive.UQueryHiveACLSessionStateBuilder") \       
.config("spark.sql.catalog.class", "org.apache.spark.sql.hive.UQueryHiveACLExternalCatalog") \  
.config("spark.sql.extensions","org.apache.spark.sql.DliSparkExtension") \
        .appName("python Spark test catalog") \
        .getOrCreate()

    spark.sql("CREATE database if not exists test_sparkapp").collect()
    spark.sql("drop table if exists test_sparkapp.dli_rds").collect()
    spark.sql(creatTbl).collect()
    spark.sql("select * from test_sparkapp.dli_rds").show()
    spark.sql("insert into table test_sparkapp.dli_rds select 12,'aaa'").collect()
    spark.sql("select * from test_sparkapp.dli_rds").show()
    spark.sql("insert overwrite table test_sparkapp.dli_rds select 1111,'asasasa'").collect()
    spark.sql("select * from test_sparkapp.dli_rds").show()
    spark.sql("drop table test_sparkapp.dli_rds").collect()
    spark.stop()