Google Cloud (Dataproc) — Artifact Registry or GCS bucket

Route A: private Artifact Registry Maven repository for normal Maven-coordinate resolution.
Route B: private Cloud Storage bucket for the ZIP/JAR bundle, using customer IAM or a short-lived signed URL.

The package targets Dataproc/Spark and includes no Python wheel. Keep both routes private and use the customer's approved service accounts, IAM, encryption, Region, and networking policy.

1. Validate and authorize the release

Verify the customer, commercial agreement/order, DataAI version, Google Cloud project, Region, Dataproc runtime, Java/Scala/Spark baseline, environments, service accounts, and support owner.
Obtain DataAIETL-Google-1.0.0.zip and ARTIFACTS.sha256. Recompute hashes and run the smoke test against the claimed runtime.
Choose Artifact Registry when builds/jobs must resolve Maven coordinates. Choose Cloud Storage when the customer attaches explicit JAR URIs or downloads the full authorized bundle.
Record artifact coordinates or object paths, version, hashes, project/Region, approver, agreement/order reference, and entitlement dates.

2A. Publish to Artifact Registry Maven

Enable Artifact Registry in the approved project and create a private Maven repository in the required Region. Use a production release repository distinct from evaluation/snapshots.
Grant the publisher only Artifact Registry write permission needed for release. Grant customer build/job identities read-only permission on the repository.
Install/configure the Google Cloud CLI and Artifact Registry Maven credential helper using Google's current authentication guidance. Prefer workload identity or short-lived credentials over service-account keys.
Add the repository URL/ID to the release Maven configuration without committing access tokens or private keys.
Deploy the approved POMs and JARs with their immutable 1.0.0 coordinates. Do not overwrite or reuse a published version for different bytes.
List the repository packages/versions and confirm all expected modules are present and anonymous/public access is absent.
With a clean Maven cache and a read-only test service account, resolve the top-level dependency, compare downloaded hashes, and run the DataAI Spark smoke test.
Give the customer the repository location, exact coordinates/version, IAM/credential-helper setup, checksums, production guides, entitlement reference, and support path.

2B. Fulfill through a private Cloud Storage bucket

Create or select a private bucket and versioned customer prefix in the approved project, Region, storage class, and encryption configuration. Enforce public-access prevention.
Upload the immutable production ZIP and checksum manifest. Use object versioning/retention according to release policy.
For ongoing use, grant the authorized customer or Dataproc service account read permission only on the needed objects/prefix.
For one-time delivery, create a V4 signed URL with the shortest practical expiration. A signed URL is a bearer credential; send it only through the approved channel and avoid putting it in logs or tickets.
Test as the customer identity or from a clean session: download the package/checksum, recompute SHA-256, extract, and attach the JARs to an isolated Dataproc batch or cluster job.
Give the customer exact gs:// URIs or signed links, object generations where used, hashes, supported runtime, IAM requirements, guides, entitlement reference, and support path.
Remove the IAM grant or allow/revoke signed access when authorization ends.

3. Customer acceptance on Dataproc

Attach the exact versioned JARs through the Dataproc job's supported --jars or dependency configuration.
Use a restricted runtime service account, private networking as required, customer-approved connectors, secret management, and non-production acceptance outputs.
Run fictional data through the packaged verification flow and record runtime version, job ID, artifact versions/hashes, output location, and acceptance contact.
Promote only the already verified immutable artifacts to production configuration.

4. Updates and revocation

Publish changed bytes under a new semantic Maven version or new versioned GCS object path/generation.
Repeat clean resolution/download, hash, Dataproc compatibility, IAM, logging, and output tests.
Revoke reader roles and signed access when commercial rights expire.

Official references

Back to DataAI ETL Markets