Apache Iceberg

Apache Iceberg source connector

Description

Source connector for Apache Iceberg. It can support batch and stream mode.

Key features

• batch
• stream
• exactly-once
• column projection
• parallelism
• support user-defined split
• data format
  • parquet
  • orc
  • avro
• iceberg catalog
  • hadoop(2.7.1 , 2.7.5 , 3.1.3)
  • hive(2.3.9 , 3.1.2)

Options

name	type	required	default value
catalog_name	string	yes	-
catalog_type	string	yes	-
uri	string	no	-
warehouse	string	yes	-
namespace	string	yes	-
table	string	yes	-
schema	config	no	-
case_sensitive	boolean	no	false
start_snapshot_timestamp	long	no	-
start_snapshot_id	long	no	-
end_snapshot_id	long	no	-
use_snapshot_id	long	no	-
use_snapshot_timestamp	long	no	-
stream_scan_strategy	enum	no	FROM_LATEST_SNAPSHOT
common-options		no	-

catalog_name [string]

User-specified catalog name.

catalog_type [string]

The optional values are:

• hive: The hive metastore catalog.
• hadoop: The hadoop catalog.

uri [string]

The Hive metastore’s thrift URI.

warehouse [string]

The location to store metadata files and data files.

namespace [string]

The iceberg database name in the backend catalog.

table [string]

The iceberg table name in the backend catalog.

case_sensitive [boolean]

If data columns where selected via schema [config], controls whether the match to the schema will be done with case sensitivity.

schema [config]

fields [Config]

Use projection to select data columns and columns order.

e.g.

schema {
    fields {
      f2 = "boolean"
      f1 = "bigint"
      f3 = "int"
      f4 = "bigint"
    }
}

start_snapshot_id [long]

Instructs this scan to look for changes starting from a particular snapshot (exclusive).

start_snapshot_timestamp [long]

Instructs this scan to look for changes starting from the most recent snapshot for the table as of the timestamp. timestamp – the timestamp in millis since the Unix epoch

end_snapshot_id [long]

Instructs this scan to look for changes up to a particular snapshot (inclusive).

use_snapshot_id [long]

Instructs this scan to look for use the given snapshot ID.

use_snapshot_timestamp [long]

Instructs this scan to look for use the most recent snapshot as of the given time in milliseconds. timestamp – the timestamp in millis since the Unix epoch

stream_scan_strategy [enum]

Starting strategy for stream mode execution, Default to use FROM_LATEST_SNAPSHOT if don’t specify any value. The optional values are:

• TABLE_SCAN_THEN_INCREMENTAL: Do a regular table scan then switch to the incremental mode.
• FROM_LATEST_SNAPSHOT: Start incremental mode from the latest snapshot inclusive.
• FROM_EARLIEST_SNAPSHOT: Start incremental mode from the earliest snapshot inclusive.
• FROM_SNAPSHOT_ID: Start incremental mode from a snapshot with a specific id inclusive.
• FROM_SNAPSHOT_TIMESTAMP: Start incremental mode from a snapshot with a specific timestamp inclusive.

common options

Source plugin common parameters, please refer to Source Common Options for details.

Example

simple

source {
  Iceberg {
    catalog_name = "seatunnel"
    catalog_type = "hadoop"
    warehouse = "hdfs://your_cluster//tmp/seatunnel/iceberg/"
    namespace = "your_iceberg_database"
    table = "your_iceberg_table"
  }
}

Or

source {
  Iceberg {
    catalog_name = "seatunnel"
    catalog_type = "hive"
    uri = "thrift://localhost:9083"
    warehouse = "hdfs://your_cluster//tmp/seatunnel/iceberg/"
    namespace = "your_iceberg_database"
    table = "your_iceberg_table"
  }
}

column projection

source {
  Iceberg {
    catalog_name = "seatunnel"
    catalog_type = "hadoop"
    warehouse = "hdfs://your_cluster/tmp/seatunnel/iceberg/"
    namespace = "your_iceberg_database"
    table = "your_iceberg_table"

    schema {
      fields {
        f2 = "boolean"
        f1 = "bigint"
        f3 = "int"
        f4 = "bigint"
      }
    }
  }
}

tip

In order to be compatible with different versions of Hadoop and Hive, the scope of hive-exec and flink-shaded-hadoop-2 in the project pom file are provided, so if you use the Flink engine, first you may need to add the following Jar packages to <FLINK_HOME>/lib directory, if you are using the Spark engine and integrated with Hadoop, then you do not need to add the following Jar packages.

flink-shaded-hadoop-x-xxx.jar
hive-exec-xxx.jar
libfb303-xxx.jar

Some versions of the hive-exec package do not have libfb303-xxx.jar, so you also need to manually import the Jar package.

Changelog

2.2.0-beta 2022-09-26

• Add Iceberg Source Connector

next version

• [Feature] Support Hadoop3.x (3046)
• [improve][api] Refactoring schema parse (4157)