Interconnecting FlinkServer with HDFS

Scenario

This section describes the data definition language (DDL) of HDFS as a sink table, as well as the WITH parameters and example code for creating a sink table, and provides guidance on how to perform operations on the FlinkServer job management page.

If your Kafka cluster is in security mode, the following example SQL statements can be used.

Prerequisites

• The HDFS, Yarn, and Flink services have been installed in a cluster.
• The client that contains the HDFS service has been installed, for example, in the /opt/Bigdata/client directory.
• You have created a user with FlinkServer Admin Privilege, for example, flink_admin, to access the Flink web UI. For details, see Authentication Based on Users and Roles.

Procedure

1. Log in to Manager as user flink_admin and choose Cluster > Services > Flink. In the Basic Information area, click the link on the right of Flink WebUI to access the Flink web UI.
2. Create a Flink SQL job by referring to Creating a Job. On the job development page, configure the job parameters as follows and start the job.
   Select Enable CheckPoint in Running Parameter and set Time Interval (ms) to 60000.

   CREATE TABLE kafka_table (
     user_id STRING,
     order_amount DOUBLE,
     log_ts TIMESTAMP(3),
     WATERMARK FOR log_ts AS log_ts - INTERVAL '5' SECOND
   ) WITH (
     'connector' = 'kafka',
     'topic' = 'user_source',
     'properties.bootstrap.servers' = 'IP address of the Kafka broker instance:Kafka port number',
     'properties.group.id' = 'testGroup',
     'scan.startup.mode' = 'latest-offset',
     'format' = 'csv',
      --Ignore the CSV data that fails to be parsed.
     'csv.ignore-parse-errors' = 'true' ,--If the data is in JSON format, set 'json.ignore-parse-errors' to true.
     'properties.sasl.kerberos.service.name' = 'kafka',
     'properties.security.protocol' = 'SASL_PLAINTEXT',
     'properties.kerberos.domain.name' = 'hadoop.System domain name'
   
   );
   
   CREATE TABLE fs_table (
     user_id STRING,
     order_amount DOUBLE,
     dt STRING,
     `hour` STRING
   ) PARTITIONED BY (dt, `hour`) WITH ( --Date-specific file partitioning
     'connector'='filesystem',
     'path'='hdfs:///sql/parquet',
     'format'='parquet',
     'sink.partition-commit.delay'='1 h',
     'sink.partition-commit.policy.kind'='success-file'
   );
   -- streaming sql, insert into file system table
   INSERT INTO fs_table SELECT user_id, order_amount, DATE_FORMAT(log_ts, 'yyyy-MM-dd'), DATE_FORMAT(log_ts, 'HH') FROM kafka_table;

   

   Kafka port number

   • In security mode, the port number is the value of sasl.port (21007 by default).
   • In non-security mode, the port is the value of port (9092 by default). If the port number is set to 9092, set allow.everyone.if.no.acl.found to true. The procedure is as follows:

     Log in to FusionInsight Manager and choose Cluster > Services > Kafka. On the displayed page, click Configurations and then All Configurations, search for allow.everyone.if.no.acl.found, set its value to true, and click Save.

3. On the job management page, check whether the job status is Running.
4. Execute the following commands to view the topic and write data to Kafka. For details, see Managing Messages in Kafka Topics.

   ./kafka-topics.sh --list --zookeeper IP address of the ZooKeeper quorumpeer instance:ZooKeeper port number/kafka

   sh kafka-console-producer.sh --broker-list IP address of the node where the Kafka instance locates:Kafka port number --topic Topic name --producer.config Client directory/Kafka/kafka/config/producer.properties

   For example, if the topic name is user_source, the script is sh kafka-console-producer.sh --broker-list IP address of the node where the Kafka instance locates:Kafka port number --topic user_source --producer.config /opt/Bigdata/client/Kafka/kafka/config/producer.properties.

   Enter the message content.

   3,3333,"2021-09-10 14:00"
   4,4444,"2021-09-10 14:01"

   Press Enter to send the message.

   

   • IP address of the ZooKeeper quorumpeer instance

     To obtain IP addresses of all ZooKeeper quorumpeer instances, log in to FusionInsight Manager and choose Cluster > Services > ZooKeeper. On the displayed page, click Instance and view the IP addresses of all the hosts where the quorumpeer instances locate.

   • Port number of the ZooKeeper client

     Log in to FusionInsight Manager and choose Cluster > Service > ZooKeeper. On the displayed page, click Configurations and check the value of clientPort. The default value is 24002.

5. Run the following command to check whether data is written from the HDFS directory to the sink table:

   hdfs dfs -ls -R /sql/parquet

Interconnecting Flink with HDFS Partitions

• Customized partitioning

  Flink's file system supports partitions in the standard Hive format. You do not need to register partitions with a table catalog. Partitions are inferred based on the directory structure.

  For example, a table that is partitioned based on the following directory is inferred to contain datetime and hour partitions.

  path
  └── datetime=2021-09-03
      └── hour=11
          ├── part-0.parquet
          ├── part-1.parquet
      └── hour=12
          ├── part-0.parquet
  └── datetime=2021-09-24
      └── hour=6
          ├── part-0.parquet

• Rolling policy of partition files

  Data in the partition directories is split into part files. Each partition contains at least one part file, which is used to receive the data written by the subtask of the sink.

  The following parameters describe the rolling policies of partition files.

  Parameter	Default Value	Type	Description
  sink.rolling-policy.file-size	128 MB	MemorySize	Maximum size of a partition file before it is rolled.
  sink.rolling-policy.rollover-interval	30 minutes	Duration	Maximum duration that a partition file can stay open before it is rolled.
  sink.rolling-policy.check-interval	1 minute	Duration	Interval for checking time-based rolling policies.

• File merging
  File compression is supported, allowing applications to have a shorter checkpoint interval without generating a large number of files.

  

  Only files in a single checkpoint are compressed. That is, the number of generated files is at least the same as the number of checkpoints. Files are invisible before merged. They are visible after both the checkpoint and compression are complete. If file compression takes too much time, the checkpoint will be prolonged.

  Parameter	Default Value	Type	Description
  auto-compaction	false	Boolean	Whether to enable automatic compression. Data will be written to temporary files. After a checkpoint is complete, the temporary files generated by the checkpoint are compressed. These temporary files are invisible before compression.
  compaction.file-size	none	MemorySize	Size of the target file to be compressed. The default value is the size of the file to be rolled.

• Partition commit
  After a file is written to a partition, for example, a partition is added to Hive metastore (HMS) or a _SUCCESS file is written to a directory, the downstream application needs to be notified. Triggers and policies are used to commit partition files.

  • Trigger parameters for committing partition files

    Parameter	Default Value	Type	Description
    sink.partition-commit.trigger	process-time	String	process-time: System time of the compute node. It does not need to extract the partition time or generate watermarks. If the current system time exceeds the system time generated when a partition is created plus the delay time, the partition should be submitted. partition-time: Time extracted from the partition. Watermarks are required. If the time for generating watermarks exceeds the time extracted from a partition plus the delay time, the partition should be submitted.
    sink.partition-commit.delay	0 s	Duration	Partitions will not be committed before the delay time. If it is a daily partition, the value is 1 d. If it is an hourly one, the value is 1 h.

  • Policy parameters for committing partition files

    Parameter	Default Value	Type	Description
    sink.partition-commit.policy.kind	-	String	Policy for committing partitions: metastore: used to add partitions to metastore. Only Hive tables support the metastore policy. The file system manages partitions based on the directory structure. success-file: used to add success-file files to a directory. The two policies can be configured at the same time, that is, 'sink.partition-commit.policy.kind'='metastore,success-file'.
    sink.partition-commit.policy.class	-	String	Class that implements partition commit policy interfaces. This parameter takes effect only in the customized submission policies.
    sink.partition-commit.success-file.name	_SUCCESS	String	File name of the success-file partition commit policy. The default value is _SUCCESS.