CROWDB provides a finite S3-compatible API for buckets, objects, and multipart uploads. This page lists the supported development surface; it does not claim complete Amazon S3 compatibility.
The interface contract is independent of deployment. The example below uses the current single-node evaluation option; deployment and lifecycle details live in the deployment guides.
Single-node container
docker run -d --name crowdb-s3 -p 127.0.0.1:9091:9091 crowdb/crowdb-s3:latestUse Admin deployment for full volume, health, recovery, and limit details. Open the Console at http://127.0.0.1:9090/ with the 9090 mapping, then run:
docker exec crowdb-s3 crowdb-monitor credentials show --format envSet AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, and AWS_DEFAULT_REGION from the printed environment values.
Supported API
- Buckets:
CreateBucket,HeadBucket,ListBuckets,DeleteBucket. A nonempty bucket cannot be deleted. - Objects:
PutObject,HeadObject,GetObject,DeleteObject. PUT replaces an existing key atomically; a missing-key delete succeeds. - Listing:
ListObjectsV2, prefix, delimiter, page limits, and continuation tokens. - Reads: byte ranges, ETags, content headers, and supported conditional requests.
- Multipart:
CreateMultipartUpload,UploadPart,ListParts,ListMultipartUploads,CompleteMultipartUpload,AbortMultipartUpload. - Authentication/integrity: SigV4, presigned request verification, Content-MD5 validation, and multipart ETags. Supported checksum and streaming modes are narrower than the full AWS surface.
Bucket/object, pagination, multipart replacement/completion/abort, bad-digest rejection, response-loss retries, concurrent requests, and restart recovery have boto3 real-storage test coverage. The container smoke suite separately verifies writes and reads after service recovery and persisted-volume restart.
Client coverage
The repository acceptance suite covers the AWS SDK for Python (boto3), AWS SDK for JavaScript, AWS SDK for Java, and AWS SDK for Go. It also runs AWS CLI and rclone workflows, including copy, multipart transfer, listing, sync, and delete cases. These are tested client paths; they do not imply complete S3 compatibility.
After the container is healthy, set the printed AWS credentials and region in your client environment. Install boto3. Use path-style addressing and the explicit checksum configuration used by the container fixture:
import boto3
from botocore.config import Config
s3 = boto3.client(
"s3",
endpoint_url="http://localhost:9091",
config=Config(
s3={"addressing_style": "path"},
request_checksum_calculation="when_required",
response_checksum_validation="when_required",
),
)
s3.create_bucket(Bucket="example")
s3.put_object(Bucket="example", Key="hello.txt", Body=b"hello")
print(s3.get_object(Bucket="example", Key="hello.txt")["Body"].read())
print(s3.list_objects_v2(Bucket="example")["Contents"])
s3.delete_object(Bucket="example", Key="hello.txt")
s3.delete_bucket(Bucket="example")
Current exclusions
- Server-side CopyObject/UploadPartCopy and batch DeleteObjects are planned.
- Versioning, lifecycle policies, ACLs/bucket policies, server-side encryption, object tags, replication, notifications, and object locking are outside the advertised surface.
- Filesystem mounting with s3fs-fuse is unsupported. Other SDK operations outside the tested workflows still require separate acceptance.
Storage and recovery semantics
Publication exposes a complete object, never partial upload bytes. Reads retain one immutable generation. Deletion removes logical visibility before reclamation. Multipart state survives restart. S3 buckets do not expose Iceberg's table files, and uploading an object does not register a table.
For bare-metal service wiring, see data-access deployment. CLI-managed mini-clusters use allocated ports printed by the CLI; the container defaults above do not change their port allocation.