How to configure S3 storage backends in GX 1.9

[Saamu192]

Hi everyone,

I recently migrated my Great Expectations project from version 1.4.4 to 1.9.0.

In my previous setup, I used S3StoreBackendDefaults to easily configure AWS S3 as the storage backend for my expectations, validation results, checkpoints, and data docs. The code looked like this:

ctx = gx.get_context(
    project_config=DataContextConfig(
        store_backend_defaults=S3StoreBackendDefaults(
            default_bucket_name="...",
            expectations_store_prefix=f"great-expectations/{context_name}/expectations/",
            validation_results_store_prefix=f"great-expectations/{context_name}/uncommitted/validations/",
            validation_definition_store_prefix=f"great-expectations/{context_name}/validation_definitions/",
            checkpoint_store_prefix=f"great-expectations/{context_name}/checkpoints/",
            data_docs_prefix=f"great-expectations/{context_name}/uncommitted/data_docs/",
        )
    ),
)

After reviewing the V1 migration guides and the latest documentation, I noticed that only ephemeral, in-memory, or local filesystem contexts seem to be natively configurable. I am struggling to find an equivalent implementation for S3 in the updated API.

My questions are:

1. Was cloud-specific store backend support (like S3/Azure/GCS) completely removed from GX Core?
2. If so, is the recommended path for cloud-native setups now exclusively centered around GX Cloud, or is there a standard way to achieve the same S3-backed behavior programmatically using GX Core 1.9.x? (e.g., manually syncing local files via boto3 or mount tools like s3fs).

I appreciate any guidance, workarounds, or clarification on the architectural direction of the library regarding cloud metadata storage!

Thank you!

[joshua-stauffer]

hi @Saamu192, thanks for the question. Yes, cloud-specific store backend support was removed in GX v1.0. The implementations were just thin wrappers around the cloud SDK. They weren't up to our standards and fell outside the core scope of the library, and so we decided to remove them from the API with the v1 release.

As you guessed, the recommended path is to sync your FileDataContext to/from the bucket with the cloud SDK each run. Pull down your gx directory, run your Checkpoint, and push back only the result under a unique key name. Two places are important to avoid concurrency: updating your GX config itself, and building DataDocs. For updating your config it's best to have a single source of truth writing the config to the bucket, such as a CI action. For DataDocs, build from a single place when you're ready to publish or consume the site.

I'm working on some documentation for this pattern, so any feedback you might have is welcome!