Skip to content

[Java IO] Add ArrowFlight IO connector#37904

Open
bvolpato wants to merge 5 commits intoapache:masterfrom
bvolpato:bvolpato/arrow-flight-io
Open

[Java IO] Add ArrowFlight IO connector#37904
bvolpato wants to merge 5 commits intoapache:masterfrom
bvolpato:bvolpato/arrow-flight-io

Conversation

@bvolpato
Copy link
Copy Markdown
Contributor

Summary

Add a new I/O connector for Apache Arrow Flight, enabling high-performance data transfer over gRPC using the Arrow columnar format. Arrow Flight avoids serialization/deserialization overhead and can achieve 10-50x faster data transfer compared to JDBC/ODBC.

Fixes #20116

Changes

  • ArrowFlightIO.Read — BoundedSource-based read transform that fetches Arrow Flight schemas via getInfo, splits work by FlightEndpoint, and converts record batches to Beam Row using the existing ArrowConversion extension.
  • ArrowFlightIO.Write — DoFn-based write transform that buffers Beam Rows into configurable batches, converts them to Arrow vectors, and streams them to a Flight server via doPut.
  • Endpoint-level split parallelism — each FlightEndpoint returned by the server becomes a separate sub-source for parallel reads.
  • Bearer token authentication — optional token support via withToken().
  • Build integration — registered flight-core and flight-sql artifacts in BeamModulePlugin.groovy; added the :sdks:java:io:arrow-flight module to settings.gradle.kts.
  • Documentation — added ArrowFlightIO to CHANGES.md (2.73.0) and the built-in connectors table in connectors.md.

Testing

Tests use a real in-process FlightServer with a TestFlightProducer:

  • testRead() — starts a Flight server serving two rows, reads them with ArrowFlightIO.read(), and asserts the output via PAssert.
  • testWrite() — writes two rows with ArrowFlightIO.write() and verifies the producer received exactly 2 records.

Both tests run the full pipeline end-to-end using TestPipeline and the DirectRunner.

Did this cause any problems?

Rollback/revert this commit and redeploy the service.


Thank you for your contribution! Follow this checklist to help us incorporate your contribution quickly and easily:

  • Mention the appropriate issue in your description (for example: addresses #123), if applicable. This will automatically add a link to the pull request in the issue. If you would like the issue to automatically close on merging the pull request, comment fixes #<ISSUE NUMBER> instead.
  • Update CHANGES.md with noteworthy changes.
  • If this contribution is large, please file an Apache Individual Contributor License Agreement.

See the Contributor Guide for more tips on how to make review process smoother.

To check the build health, please visit https://github.com/apache/beam/blob/master/.test-infra/BUILD_STATUS.md

GitHub Actions Tests Status (on master branch)

Build python source distribution and wheels
Python tests
Java tests
Go tests

See CI.md for more information about GitHub Actions CI or the workflows README to see a list of phrases to trigger workflows.

@gemini-code-assist
Copy link
Copy Markdown
Contributor

Summary of Changes

Hello, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request introduces a new Java I/O connector for Apache Arrow Flight, significantly enhancing data transfer capabilities within Beam pipelines. The connector leverages Arrow Flight's high-performance RPC framework to facilitate efficient reading from and writing to Arrow Flight servers, minimizing serialization overhead and enabling faster data movement. This integration expands Beam's ecosystem by providing a robust solution for interacting with various Arrow Flight-compatible data systems.

Highlights

  • ArrowFlightIO.Read: Implemented a BoundedSource-based read transform that fetches Arrow Flight schemas, splits work by FlightEndpoint, and converts record batches to Beam Rows using the existing ArrowConversion extension.
  • ArrowFlightIO.Write: Developed a DoFn-based write transform that buffers Beam Rows into configurable batches, converts them to Arrow vectors, and streams them to a Flight server via doPut.
  • Endpoint-level Split Parallelism: Enabled each FlightEndpoint returned by the server to become a separate sub-source for parallel reads, enhancing data ingestion efficiency.
  • Bearer Token Authentication: Added optional support for bearer token authentication via withToken() for secure connections.
  • Build Integration: Registered flight-core and flight-sql artifacts in BeamModulePlugin.groovy and added the new :sdks:java:io:arrow-flight module to settings.gradle.kts.
  • Documentation: Updated CHANGES.md and the built-in connectors table in connectors.md to reflect the addition of the ArrowFlightIO connector.

🧠 New Feature in Public Preview: You can now enable Memory to help Gemini Code Assist learn from your team's feedback. This makes future code reviews more consistent and personalized to your project's style. Click here to enable Memory in your admin console.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature Command Description
Code Review /gemini review Performs a code review for the current pull request in its current state.
Pull Request Summary /gemini summary Provides a summary of the current pull request in its current state.
Comment @gemini-code-assist Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help /gemini help Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for GitHub and other Google products, sign up here.

Footnotes

  1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution.

@bvolpato bvolpato force-pushed the bvolpato/arrow-flight-io branch 7 times, most recently from b384457 to 1b3518d Compare March 22, 2026 21:20
@bvolpato bvolpato marked this pull request as ready for review March 22, 2026 21:43
@github-actions
Copy link
Copy Markdown
Contributor

Assigning reviewers:

R: @chamikaramj for label java.
R: @damccorm for label build.
R: @shunping for label website.

Note: If you would like to opt out of this review, comment assign to next reviewer.

Available commands:

  • stop reviewer notifications - opt out of the automated review tooling
  • remind me after tests pass - tag the comment author after tests pass
  • waiting on author - shift the attention set back to the author (any comment or push by the author will return the attention set to the reviewers)

The PR bot will only process comments in the main thread (not review comments).

Copy link
Copy Markdown
Contributor

@Abacn Abacn left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks for the new IO! Had an initial comment to kick off

Comment thread settings.gradle.kts
@bvolpato bvolpato force-pushed the bvolpato/arrow-flight-io branch from e1e26fd to d0551f0 Compare April 2, 2026 13:09
bvolpato added 2 commits April 5, 2026 22:22
Add a new IO connector for Apache Arrow Flight, enabling high-performance
data transfer over gRPC using the Arrow columnar format.

Includes read (BoundedSource) and write (DoFn with doPut) support with
endpoint-level split parallelism and bearer token authentication.

Fixes apache#20116
- Register :sdks:java:io:arrow-flight in javaioPreCommit task (build.gradle.kts)
- Add --add-opens JVM arg for Arrow native memory on JDK 17+
- Make host() nullable at AutoValue level to fix factory method NPE
- Eagerly materialize rows from Arrow buffers to prevent stale access
- Move root.setRowCount() after vector population for correct ordering
- Use AtomicInteger for thread-safe write record counting in tests
@bvolpato bvolpato force-pushed the bvolpato/arrow-flight-io branch from d0551f0 to 7252108 Compare April 6, 2026 02:22
@bvolpato
Copy link
Copy Markdown
Contributor Author

Run Java_Amazon-Web-Services2_IO_Direct PreCommit

@github-actions github-actions Bot added aws and removed aws labels Apr 12, 2026
@github-actions
Copy link
Copy Markdown
Contributor

Reminder, please take a look at this pr: @chamikaramj @damccorm @shunping

break;
default:
throw new IllegalArgumentException(
String.format(
Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can we add a TODO here about supporting aggregation and logical types? Currently supported types are pretty limited

}

test {
jvmArgs '--add-opens=java.base/java.nio=ALL-UNNAMED'
Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I understsnd add-opens are needed:

Caused by: java.lang.ExceptionInInitializerError: Exception java.lang.RuntimeException: Failed to initialize MemoryUtil. You must start Java with `--add-opens=java.base/java.nio=org.apache.arrow.memory.core,ALL-UNNAMED` (See https://arrow.apache.org/docs/java/install.html) [in thread "direct-runner-worker"]
	at org.apache.arrow.memory.util.MemoryUtil.<clinit>(MemoryUtil.java:140)

this means using ArrowFlightIO one needs to set --JdkAddOpenModules=java.base/java.nio=ALL-UNNAMED pipeline options in Java17+. We should note this here and in ArrowFlightIO Javadoc

Also, Java8 runtime doesn't recognize '--add-opens' pipeline options. However we'll soon remove Java8 support so it's fine here.

Read spec = source.spec;

String hostName = checkNotNull(spec.host(), "host");
if (source.endpoint != null) {
Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

A potential issue arises when a source (with endPoint=null) first get read (from endpoint[0]) then split is called on getSource(), it will produce a list of sources for each available endpoint, because source.endpoint=null here and source.split() will unconditionally split. Would this result in any issue like duplicate records get read?

Note that source are usually split before read, with a few exceptions (e.g. BoundedReadFromUnbounded fallback scenario)

return ArrowType.Binary.INSTANCE;
case DATETIME:
return new ArrowType.Timestamp(TimeUnit.MILLISECOND, "UTC");
default:
Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

same here, add TODO for aggregation and logical types

}

@Override
public PDone expand(PCollection<Row> input) {
Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

PDone is discouraged. Please add a TODO about implementing dead letter queue (preferred) or WriteWithResult

return new Field(beamField.getName(), fieldType, Collections.emptyList());
}

private ArrowType beamTypeToArrowType(Schema.FieldType beamType) {
Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Consider move Beam schema->Arrow Schema to ArrowConversion as a public API (the inversion of toBeamSchema)

@github-actions
Copy link
Copy Markdown
Contributor

Assigning new set of reviewers because Pr has gone too long without review. If you would like to opt out of this review, comment assign to next reviewer:

R: @kennknowles for label java.
R: @Abacn for label build.
R: @kennknowles for label website.

Available commands:

  • stop reviewer notifications - opt out of the automated review tooling
  • remind me after tests pass - tag the comment author after tests pass
  • waiting on author - shift the attention set back to the author (any comment or push by the author will return the attention set to the reviewers)

@bvolpato
Copy link
Copy Markdown
Contributor Author

bvolpato commented May 5, 2026

[review-prs-verification]
Head: 3e4fe51
Ready to merge: no

Files reviewed:

  • CHANGES.md
  • build.gradle.kts
  • buildSrc/src/main/groovy/org/apache/beam/gradle/BeamModulePlugin.groovy
  • sdks/java/io/amazon-web-services2/src/test/java/org/apache/beam/sdk/io/aws2/sqs/SqsIOWriteBatchesTest.java
  • sdks/java/io/arrow-flight/build.gradle
  • sdks/java/io/arrow-flight/src/main/java/org/apache/beam/sdk/io/arrowflight/ArrowFlightIO.java
  • sdks/java/io/arrow-flight/src/main/java/org/apache/beam/sdk/io/arrowflight/package-info.java
  • sdks/java/io/arrow-flight/src/test/java/org/apache/beam/sdk/io/arrowflight/ArrowFlightIOTest.java
  • settings.gradle.kts
  • website/www/site/content/en/documentation/io/connectors.md

Tests/checks run and results:

  • JAVA_HOME=$HOME/.sdkman/candidates/java/21.0.6-tem ./gradlew --no-daemon --console=plain :sdks:java:io:arrow-flight:spotlessCheck :sdks:java:io:arrow-flight:test :sdks:java:io:amazon-web-services2:test --tests org.apache.beam.sdk.io.aws2.sqs.SqsIOWriteBatchesTest ✅ before rebase
  • Same targeted Gradle command ✅ after rebase/merge conflict resolution
  • GitHub checks snapshot as of 2026-05-05T09:02:32Z: 24 success, 1 neutral, 1 in progress, 0 failures
  • Remaining in-progress check: beam_PreCommit_Java_GCP_IO_Direct (Run Java_GCP_IO_Direct PreCommit)
  • gh pr checks --required reports no required checks for this branch

Fixes committed/pushed:

  • Merged origin/master into bvolpato/arrow-flight-io
  • Resolved CHANGES.md conflict by keeping both ArrowFlightIO and upstream IcebergIO release-note entries
  • Pushed merge commit 3e4fe5100645397449b062ceebd3aeaa218162d6 to bvolpato/arrow-flight-io

Final CI status:

  • Not final yet: one optional Beam precommit check is still running and has not exposed logs while in progress.
  • PR is now mergeable/clean from GitHub mergeability perspective, but status is unstable while that check runs.

Merge/close result:

  • Not merged or closed from this pass because final optional CI is still pending.

@bvolpato
Copy link
Copy Markdown
Contributor Author

bvolpato commented May 5, 2026

[review-prs-verification]
Ready to merge: yes

Verified after resolving the master merge conflict and pushing 3e4fe5100645397449b062ceebd3aeaa218162d6:

  • ./gradlew --no-daemon --console=plain :sdks:java:io:arrow-flight:spotlessCheck :sdks:java:io:arrow-flight:test :sdks:java:io:amazon-web-services2:test --tests org.apache.beam.sdk.io.aws2.sqs.SqsIOWriteBatchesTest
  • GitHub checks on PR head: 25 success, 1 neutral, 0 pending, 0 failing.

Notes:

  • Conflict was only in CHANGES.md; kept ArrowFlight entry plus upstream Iceberg entries.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Add an Arrow Flight based Java IO connector

2 participants