> ## Documentation Index
> Fetch the complete documentation index at: https://docs.tensor9.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Quickstart: Google Cloud to AWS
> Send Google Cloud Storage requests through Cloud Adapter to Amazon S3.
Keep the Google Cloud Storage client and request model while Cloud Adapter stores objects in Amazon S3.
## Compatibility snapshot
The [Cloud Storage to AWS profile](/cloud-adapter/service-catalog/google-cloud/databases-storage/gcs#on-aws) supports core JSON object CRUD and listing, XML request re-signing, lifecycle, and versioning behavior. Google retention policies and notifications are outside the profile. Generation identifiers adapt to S3 version or ETag behavior. `ARCHIVE` and `COLDLINE` semantics differ because some S3 objects require restore before read, and compose rejects non-final parts smaller than 5 MiB.
Existing object bodies, versions, and shared links are not copied automatically.
## Prerequisites
* an installed Google Cloud-origin Cloud Adapter endpoint;
* a Cloud Storage to S3 service adapter;
* an origin-side Google credential accepted by the endpoint;
* an AWS runtime identity with access to one test bucket;
* a separate read-only AWS verification profile;
* Python 3 with `google-cloud-storage` and AWS CLI v2.
## Set values
```bash theme={null}
export ADAPTER_STORAGE_ENDPOINT="https://adapter.example.test"
export GOOGLE_CLOUD_PROJECT="origin-project"
export ORIGIN_BUCKET="adapter-smoke-test"
export TARGET_BUCKET="adapter-smoke-test-aws"
export AWS_VERIFY_PROFILE="aws-target-verify"
export AWS_TARGET_REGION="us-east-1"
export TEST_OBJECT="quickstart/gcp-aws-$(date +%s).txt"
```
The API endpoint shape and origin authentication in this sample are deployment-specific placeholders. Use the exact endpoint and caller credential mode supplied by your Cloud Adapter deployment. Target AWS credentials remain adapter-owned.
## Write through the Cloud Storage client
```python theme={null}
import os
import google.auth
from google.api_core.client_options import ClientOptions
from google.cloud import storage
credentials, _ = google.auth.default()
client = storage.Client(
project=os.environ["GOOGLE_CLOUD_PROJECT"],
credentials=credentials,
client_options=ClientOptions(api_endpoint=os.environ["ADAPTER_STORAGE_ENDPOINT"]),
)
bucket = client.bucket(os.environ["ORIGIN_BUCKET"])
blob = bucket.blob(os.environ["TEST_OBJECT"])
expected = b"hello from Cloud Storage to S3\n"
blob.upload_from_string(expected, content_type="text/plain")
blob.reload()
actual = blob.download_as_bytes()
assert actual == expected
print({"generation": blob.generation, "etag": blob.etag, "size": blob.size})
```
Keep endpoint construction in one client factory. Target AWS credentials do not belong in the Google Cloud application.
## Verify S3 natively
```bash theme={null}
aws s3api head-object \
--bucket "$TARGET_BUCKET" --key "$TEST_OBJECT" \
--profile "$AWS_VERIFY_PROFILE" --region "$AWS_TARGET_REGION"
aws s3api get-object \
--bucket "$TARGET_BUCKET" --key "$TEST_OBJECT" \
--profile "$AWS_VERIFY_PROFILE" --region "$AWS_TARGET_REGION" \
/tmp/cloud-adapter-target.txt
printf 'hello from Cloud Storage to S3\n' | cmp - /tmp/cloud-adapter-target.txt
```
Confirm content type and bytes. Do not copy a Google generation into code that expects an S3 version ID without following the documented identity translation.
## Exercise a failure
Request an object name that does not exist and retain the Google API error code, HTTP status, request ID, and diagnostic headers. Also test a stale generation precondition if your application uses optimistic concurrency.
## Clean up
```python theme={null}
blob.delete()
```
Confirm that the current target object is no longer readable. Versioned or soft-delete-enabled targets can retain prior versions or recovery state; use an unversioned disposable target for this quickstart or remove the created version explicitly. Bucket deletion and retained target data have separate ownership.
## Before production
Test listing and pagination, generation preconditions, compose part sizes, lifecycle, versioning, storage classes, retries, and uncertain writes. Plan existing data and shared-link migration separately. If the workload uses retention or notifications, do not assume this route supplies them.
Read the full [Cloud Storage profile](/cloud-adapter/service-catalog/google-cloud/databases-storage/gcs) and add the lifecycle to [CI/CD](/cloud-adapter/guides/ci-cd).