DLI Job Development Process

• To manage fine-grained permissions for your DLI resources using IAM, create an IAM user and grant them permissions to DLI if you are an enterprise user. For details, see Creating an IAM User and Granting Permissions.
• When using DLI for the first time, you need to update the DLI agency according to the console's guidance so that DLI can use other cloud services and perform resource O&M operations on your behalf. The agency includes permissions to obtain IAM user information, access and use VPCs, CIDR blocks, routes, and peering connections, and send notifications via SMN in case of job execution failure.

  For more information on the specific permissions included in the agency, refer to Configuring DLI Agency Permissions.

• Before submitting a job using DLI, you need to create an elastic resource pool and create queues within it. This will provide the necessary compute resources for running the job. For how to create an elastic resource pool and create queues within it, see Overview of DLI Elastic Resource Pools and Queues.

  Alternatively, you can enhance DLI's computing environment by creating custom images. Specifically, to enhance the functions and performance of Spark and Flink jobs, you can create custom images by downloading the base images provided by DLI and adding dependencies (files, JAR files, or software) and private capabilities required for job execution. This changes the container runtime environment for the jobs.

  For example, you can add a Python package or C library related to machine learning to a custom image to help you extend functions. For how to create a custom image, see Enhancing the Job Runtime Environment Using a Custom Image.

• DLI metadata is the basis for developing SQL and Spark jobs. Before executing a job, you need to define databases and tables based on your business scenario.
  Note

  Flink allows for dynamic data types, enabling the definition of data structures at runtime without the need for predefined metadata.

  • Define your data structures, including data catalogs, databases, and tables. For details, see Creating a Data Directory, Database, and Table.
  • Create a bucket to store temporary data generated during job running, such as job logs and job results. For details, see Configuring a DLI Job Bucket.
  • Configure the permission to access metadata. For details, see Configuring Database Permissions on the DLI Console and Configuring Table Permissions on the DLI Console.

• DLI allows you to analyze and query data stored in OBS without the need to migrate it. Simply upload your data to OBS and use DLI for data analysis.
• Cross-source access can reduce data replication and latency when real-time access and processing of data from different sources is required for service needs.

  The prerequisites for cross-source access are that DLI can communicate with the data source network and DLI can obtain the access credentials to the data source.

  • Configure network connection between DLI and the data source by referring to Configuring the Network Connection Between DLI and Data Sources (Enhanced Datasource Connection).
  • Manage data source credentials.
    • You can use DLI's datasource authentication to manage the authentication information for accessing a specified datasource.

      This applies to SQL jobs and Flink 1.12 jobs. For details, see Using DLI Datasource Authentication to Manage Access Credentials for Data Sources.

    • You can also use DEW to manage access credentials for data sources and use a custom agency to authorize DLI to access DEW.

      This applies to Spark 3.3.1 or later and Flink 1.15 or later.

      For details, see Managing Data Source Access Credentials Using DEW and Configuring an Agency to Allow DLI to Access Other Cloud Services.

• DLI offers a serverless service that integrates stream processing, batch processing, and interactive analytics. It supports various job types to meet different data processing needs.

  Table 1 Job types supported by DLI

  Job Type	Description	Use Case
  SQL job	This type is suitable for scenarios where standard SQL statements are used for querying. It is typically used for querying and analyzing structured data. For details, see Creating and Submitting a SQL Job.	It applies to scenarios such as data warehouse query, report generation, and online analytical processing (OLAP).
  Flink job	This type is specifically designed for real-time data stream processing, making it ideal for scenarios that require low latency and quick response. It is well-suited for real-time monitoring and online analysis. Flink OpenSource job: DLI provides standard connectors and various APIs to facilitate quick integration with other data systems. For details, see Creating a Flink OpenSource SQL Job. Flink Jar job: allows you to submit Flink jobs compiled into JAR files, providing greater flexibility and customization capabilities. It is suitable for complex data processing scenarios that require user-defined functions (UDFs) or specific library integration. The Flink ecosystem can be utilized to implement advanced stream processing logic and status management. For details, see Creating a Flink Jar Job.	It applies to scenarios that require quick response, such as real-time data monitoring and real-time recommender systems. Flink Jar jobs are suitable for data analysis scenarios that require custom stream processing logic, complex state management, or integration with specific libraries.
  Spark job	Compute jobs can be submitted through interactive sessions or batch processing. Jobs are submitted to queues created within an elastic resource pool, simplifying resource management and job scheduling. It supports multiple data sources and formats, providing rich data processing capabilities, including but not limited to SQL queries and machine learning. For details, see Creating a Spark Job.	It is suitable for large-scale data processing and analysis, such as machine learning training, log analysis, and large-scale data mining.

• Manage program packages of Jar jobs.

  DLI allows you to submit Flink or Spark jobs compiled as JAR files, which contain the necessary code and dependency information for executing the job. These files are used for specific data processing tasks such as data query, analysis, and machine learning. You can manage program packages required for jobs on the DLI console.

  To submit a Spark Jar or Flink Jar job, you must first upload the program package to OBS, create a program package in DLI, and then submit the program package, data, and job parameters to run the job. For details, see Managing Program Packages of Jar Jobs.

  Note

  For Spark 3.3.1 or later and Flink 1.15 or later, when creating a Jar job, you can directly configure the program package in OBS. Program packages cannot be read from DLI.

You can query DLI monitoring metrics and alarms through Cloud Eye management console or APIs.

For example, you can monitor the resource usage and job status of a DLI queue. For details about DLI metrics, see Monitoring DLI Using Cloud Eye.

With CTS, you can log operations related to DLI, making it easier to search, audit, and trace in the future. For the supported operations, see Using CTS to Audit DLI.