Skip to main content

Installation

Choose the Paimon connector that matches your Flink minor version. This branch contains connectors for Flink 1.16–1.20 and 2.0–2.2. Use the documentation for your Paimon release when installing a released connector.

Prerequisites

Install a Flink distribution and a Java version supported by that distribution. In the commands below, <FLINK_HOME> is the extracted Flink directory. The local quick start uses Paimon's built-in local filesystem; distributed jobs need shared storage and the dependencies listed below.

Choose an Artifact

ArtifactUse it forHow to use it
paimon-flink-<flink-version>-2.2-SNAPSHOT.jarFlink SQL, Table API, and DataStream reads and writesPut the matching bundled connector in Flink's lib directory.
paimon-flink-action-2.2-SNAPSHOT.jarCompaction, ingestion, and other command-line actionsSubmit it with flink run; see Action Jars.

Install one matching Paimon bundled connector in each Flink process. Restart existing Flink processes after changing the jars. Snapshot links below point to artifact directories; choose the timestamped jar available there. A directory may be empty until that version is published.

VersionTypeJar
Flink 2.2Bundled Jarpaimon-flink-2.2-2.2-SNAPSHOT.jar
Flink 2.1Bundled Jarpaimon-flink-2.1-2.2-SNAPSHOT.jar
Flink 2.0Bundled Jarpaimon-flink-2.0-2.2-SNAPSHOT.jar
Flink 1.20Bundled Jarpaimon-flink-1.20-2.2-SNAPSHOT.jar
Flink 1.19Bundled Jarpaimon-flink-1.19-2.2-SNAPSHOT.jar
Flink 1.18Bundled Jarpaimon-flink-1.18-2.2-SNAPSHOT.jar
Flink 1.17Bundled Jarpaimon-flink-1.17-2.2-SNAPSHOT.jar
Flink 1.16Bundled Jarpaimon-flink-1.16-2.2-SNAPSHOT.jar
Flink ActionAction Jarpaimon-flink-action-2.2-SNAPSHOT.jar

Configure Dependencies

cp paimon-flink-1.20-2.2-SNAPSHOT.jar <FLINK_HOME>/lib/

The warehouse filesystem and catalog determine the additional dependencies:

DependencyWhen it is neededSetup
Hadoop librariesHadoop-backed filesystem accessExpose installed Hadoop libraries through HADOOP_CLASSPATH, or use the Paimon Hadoop uber jar.
Filesystem connectorHDFS or object storage such as S3 and OSSFollow Filesystems for the chosen storage.
Hive connectorHive-backed catalogsFollow Creating Hive Catalog.
JDBC driverJDBC-backed catalogsFollow Creating JDBC Catalog.

For a machine with Hadoop installed:

export HADOOP_CLASSPATH="$(hadoop classpath)"

Set this before starting Flink processes and the SQL client. Use a warehouse on shared storage for a distributed cluster. A file:/tmp/... warehouse is suitable for the single-machine tutorial.

Build from Source

Clone the Paimon repository, then build the connector and its dependencies from the repository root. Select the profile for your Flink major version:

# Flink 1.20 example
mvn -pl paimon-flink/paimon-flink-1.20 -am -Pflink1 -DskipTests install

# Flink 2.2 example
mvn -pl paimon-flink/paimon-flink-2.2 -am -Pflink2 -DskipTests install

The bundled jar is written to paimon-flink/paimon-flink-<flink-version>/target/paimon-flink-<flink-version>-2.2-SNAPSHOT.jar. To build the action jar, use -pl paimon-flink/paimon-flink-action with the corresponding profile. Its output is in paimon-flink/paimon-flink-action/target/.

Continue with Quick Start to create, write, and query a table.