Streaming Write to Apache Iceberg with Lakehouse REST Catalog

The Dataflow managed I/O connector lets you ingest real-time data into Apache Iceberg tables through a REST catalog. This approach provides low-latency streaming writes and dynamic destination routing while maintaining open table format interoperability across your lakehouse analytics stack.

Managed I/O supports multiple catalog backends, including Hadoop, Hive, and BigQuery metastore, but this document focuses on streaming pipelines that target a Lakehouse REST catalog. If you write to BigQuery tables for Apache Iceberg, use the BigQueryIO connector with the BigQuery Storage API instead, as target tables must exist before ingestion.

Prerequisites

Set up Lakehouse. Configure your Google Cloud project with the required permissions by following Use the Lakehouse runtime catalog with the Iceberg REST catalog. Make sure that you understand the limitations of Lakehouse Iceberg REST Catalog described on that page.

Dependencies

Add the following dependencies to your project:

Java 

<dependency>
  <groupId>org.apache.beam</groupId>
  <artifactId>beam-sdks-java-managed</artifactId>
  <version>${beam.version}</version>
</dependency>

<dependency>
  <groupId>org.apache.beam</groupId>
  <artifactId>beam-sdks-java-io-iceberg</artifactId>
  <version>${beam.version}</version>
</dependency>

<dependency>
  <groupId>org.apache.iceberg</groupId>
  <artifactId>iceberg-gcp</artifactId>
  <version>${iceberg.version}</version>
</dependency>

Example

The following example demonstrates a streaming pipeline that writes data to an Apache Iceberg table using the REST catalog, backed by the Lakehouse runtime catalog.

Java

To authenticate to Dataflow, set up Application Default Credentials. For more information, see Set up authentication for a local development environment. 

import com.google.auth.oauth2.GoogleCredentials;
import com.google.common.collect.ImmutableMap;
import java.io.IOException;
import java.util.Map;
import org.apache.beam.sdk.Pipeline;
import org.apache.beam.sdk.coders.RowCoder;
import org.apache.beam.sdk.extensions.gcp.options.GcpOptions;
import org.apache.beam.sdk.io.GenerateSequence;
import org.apache.beam.sdk.managed.Managed;
import org.apache.beam.sdk.options.Default;
import org.apache.beam.sdk.options.Description;
import org.apache.beam.sdk.options.PipelineOptionsFactory;
import org.apache.beam.sdk.options.StreamingOptions;
import org.apache.beam.sdk.options.Validation;
import org.apache.beam.sdk.schemas.Schema;
import org.apache.beam.sdk.transforms.MapElements;
import org.apache.beam.sdk.values.Row;
import org.apache.beam.sdk.values.TypeDescriptors;
import org.joda.time.Duration;

/**
 * A streaming pipeline that writes data to an Iceberg table using the REST catalog.
 *
 * <p>This example demonstrates writing to an Iceberg table backed by the BigLake Metastore. For
 * more information on BigLake, see the documentation at
 * https://cloud.google.com/bigquery/docs/blms-rest-catalog.
 */
public class ApacheIcebergRestCatalogStreamingWrite {

  // The schema for the generated records.
  public static final Schema SCHEMA =
      Schema.builder().addStringField("user_id").addInt64Field("click_count").build();

  /** Pipeline options for this example. */
  public interface Options extends GcpOptions, StreamingOptions {
    @Description(
        "Warehouse location where the table's data will be written to. "
            + "BigLake only supports Single Region buckets")
    @Validation.Required
    String getWarehouse();

    void setWarehouse(String warehouse);

    @Description("The URI for the REST catalog")
    @Validation.Required
    @Default.String("https://biglake.googleapis.com/iceberg/v1beta/restcatalog")
    String getCatalogUri();

    void setCatalogUri(String value);

    @Description("The name of the table to write to")
    @Validation.Required
    String getIcebergTable();

    void setIcebergTable(String value);

    @Description("The name of the Apache Iceberg catalog")
    @Validation.Required
    String getCatalogName();

    void setCatalogName(String catalogName);
  }

  /**
   * The main entry point for the pipeline.
   *
   * @param args Command-line arguments
   * @throws IOException If there is an issue with Google Credentials
   */
  public static void main(String[] args) throws IOException {
    Options options = PipelineOptionsFactory.fromArgs(args).withValidation().as(Options.class);
    options.setStreaming(true);

    // Note: The token expires in 1 hour. Users may need to re-run the pipeline.
    // Future updates to Iceberg and the BigLake Metastore will support token refreshing.
    Map<String, String> catalogProps =
        ImmutableMap.<String, String>builder()
            .put("type", "rest")
            .put("uri", options.getCatalogUri())
            .put("warehouse", options.getWarehouse())
            .put("header.x-goog-user-project", options.getProject())
            .put(
                "header.Authorization",
                "Bearer "
                    + GoogleCredentials.getApplicationDefault()
                        .createScoped("https://www.googleapis.com/auth/cloud-platform")
                        .refreshAccessToken()
                        .getTokenValue())
            .put("rest-metrics-reporting-enabled", "false")
            .build();

    Map<String, Object> icebergWriteConfig =
        ImmutableMap.<String, Object>builder()
            .put("table", options.getIcebergTable())
            .put("catalog_properties", catalogProps)
            .put("catalog_name", options.getCatalogName())
            .put("triggering_frequency_seconds", 20)
            .build();

    Pipeline p = Pipeline.create(options);

    p.apply(
            "GenerateSequence",
            GenerateSequence.from(0).withRate(1, Duration.standardSeconds(5)))
        .apply(
            "ConvertToRows",
            MapElements.into(TypeDescriptors.rows())
                .via(
                    i ->
                        Row.withSchema(SCHEMA)
                            .withFieldValue("user_id", "user-" + (i % 10))
                            .withFieldValue("click_count", i % 100)
                            .build()))
        .setCoder(RowCoder.of(SCHEMA))
        .apply("WriteToIceberg", Managed.write(Managed.ICEBERG).withConfig(icebergWriteConfig));

    p.run();
  }
}

What's next