S3 handles the requests coming from S3 clients. It is based on the current amazon-S3 documentation.
S3 consists of multiple daemons listening RESTfully to http requests on a single port. We then route the requests depending on the HTTP verbs used by the request. Parsing the header allows us to determine the request type. From there, authentication will be confirmed by the Vault module using v2 or v4 authentication depending on whether the client sent a v2 or v4 authentication header. Applicable bucket and object metadata will be pulled from the metadata backend, IM-Metadata to check proper authorization (ACL's, IAM and bucket policies). If a user has been authenticated and is authorized to write/read data, we then proceed to send the write/read request to our storage backend (RING or IM-Data), before sending a response to the request sender. Any problem that arises during the handling of the request due to client error or system error will result in an error being returned to the client that follows S3's error specifications.
The multi-daemon architecture allows us to restart daemons on the fly in case of any crash without interrupting the service. The daemon handles remaining requests even in case of an error, stopping listening while another daemon is spawned to handle future requests in its stead.
Right now, the following operations are implemented:
- PutBucket
- GetBucket
- HeadBucket
- DeleteBucket
- PutBucketACL
- GetBucketACL
- PutBucketCors
- DeleteBucketCors
- GetBucketCors
- PutObject
- PutObject - Copy
- GetObject
- HeadObject
- DeleteObject
- Multi-Object Delete
- PutObjectACL
- GetObjectACL
- Multipart Upload
- Upload Part
- Upload Part - Copy
- GetService
- Put Bucket Website
- Get Bucket Website
- Delete Bucket Website
- Put Bucket Versioning
- Get Bucket Versioning
- v2 Authentication
- v4 Authentication (Transferring Payload in a Single Chunk)
- v4 Authentication (Transferring Payload in Multiple Chunks)