A microservice to move files from S3 APIs (Swift or Ceph) to other S3 APIs.

pip install s3s3==0.1.10



s3s3 is a microservice to move files from one S3 compliant service to another (Swift, Ceph, AWS).

s3s3 has two services. s3s3.scripts.listen subscribes and listens to a redis pubsub channel waiting for s3 object names. When notified it calls s3s3.api.upload and uploads that S3 object from the source to destination(s). s3s3.scripts.bucket uploads all keys in one s3 bucket to another.

s3s3.scripts.listen is meant to be a real-time daemon.

s3s3.scripts.bucket is meant to run periodically as a cron job. Anything missed by s3s3.scripts.listen will be handled by this service.


s3s3 is required because Ceph and Swift are not feature complete with AWS S3. Many available libraries that work well with AWS S3 do not work with Ceph and Swift.

Some examples: key.size and key.md5 do not work with Ceph S3 without fetching the contents of the key (or s3 object). Multipart uploads are not reliable with Ceph S3. V4 signatures are not supported by Ceph S3.


s3s3 must be configured. Configuration can be found in the s3s3.config module. A source, destination and pubsub section are required. Multiple destination connections are supported but there was minimal testing. To signify a section is a destination connection it must start with 'dest'.

Example template: https://github.com/jmatt/s3s3/blob/master/s3s3/s3s3.ini.dist

     aws_access_key_id = {YOUR_AWS_ACCESS_KEY_ID}
     aws_secret_access_key = {YOUR_AWS_SECRET_ACCESS_KEY}
     bucket_name = {YOUR_S3_BUCKET}
     host = {YOUR_CEPH_S3_ENDPOINT}
     verify_md5 = False # Verify md5 during s3 operations.
     is_secure = True # Optional
     calling_format = OrdinaryCallingFormat # Optional
     aws_access_key_id = {YOUR_AWS_ACCESS_KEY_ID}
     aws_secret_access_key = {YOUR_AWS_SECRET_ACCESS_KEY}
     bucket_name = {YOUR_S3_BUCKET}
     verify_md5 = False # Verify md5 during s3 operations.
     is_secure = True # Optional


s3s3 requires python3 and redis. It was tested with python3.4 and python3.5. And redis 3.x.

pip install s3s3


There are two clients. One for each service. s3s3.client.ListenClient is a client to listen to the 'backup' redis pubsub channel and call s3s3.client.on_notify. s3s3.client.BucketClient is a client that uses the configuration to provide access to the duplicate_bucket API function.

Command Line

Both clients are available as command line scripts.

s3s3listen --config /path/to/s3s3.ini

This will use the configuration to build source and destination boto connections, connect to redis and start listening on the backup channel. Any messages pushed to that channel will be considered source s3 key names and will attempt to be uploaded to the destination connection(s).

s3s3bucket --config /path/pto/s3s3.ini

This will use the configuration to build source and destination boto connections and duplicate the source bucket in the destination bucket.


The API can be found in s3s3.api module.

def create_connection(connection_args):

Creates a boto connection from the connection_args dictionary.

def upload(source_key, dest_key, verify_md5=False):

Upload the source key (S3 object) to the destination key. If verify_md5 is true then verify md5s match.

def duplicate_bucket(source_bucket, dest_bucket, verify_md5=False):

Duplicate the source bucket to the destination bucket. If verify_md5 is true then verify md5s match. If the md5 is not available compute it and verify it matches.


s3s3 requires redis, python3 and supervisord.

mkdir -p /opt/env
cd /opt/env
virtualenv -p python3 s3s3
. /opt/env/s3s3/bin/activate
pip install s3s3
echo_s3s3_supervisord_conf > /etc/supervisor/conf.d/s3s3.conf
echo_s3s3_ini_template > /usr/local/etc/s3s3.ini
# Update ini file with your source and destination s3 information.
service supervisor restart # or... start if it's not running.


See the LICENSE file.