Skip to content

S3 Buckets#

Before you start#

S3 is a simple object storage service initially provided by Amazon. There are numerous datasets that you can access through S3 as well as store your own objects. EPFL is providing an official S3 endpoint as well as other faculties at EPFL.

Buckets @EPFL#

At EPFL, you can choose between two S3 ressources:

Accessing S3 Buckets in the SCITAS Buckets#

You can use one of this tools to access a S3 Bucket:

s3fs#

s3fs allows you to use an S3 bucket as a filesystem. It is installed on the cluster login nodes (must be mounted before running the jobs on the login node).

Configure access to your existing S3 bucket#

echo S3_ACCESS_KEY:S3_SECRET_KEY > ${HOME}/.passwd-s3fs
chmod 600 ${HOME}/.passwd-s3fs

Mount the bucket as a filesystem in your home directory#

For EPFL S3 service, if you use an other endpoint you have to modify the url (or remove it from Amazon S3) 

mkdir ${HOME}/mybucket
s3fs BUCKET_ID ${HOME}/mybucket -o url=https://s3.epfl.ch/ -o passwd_file=${HOME}/.passwd-s3fs

Umount the filesystem 

fusermount -u ${HOME}/mybucket

boto (python)#

Install the library 

module load gcc python
pip3 install --user boto

Using the library (source: https://icitdocs.epfl.ch/display/clusterdocs/Accessing+Datasets) 

#!/usr/bin/env python3

import boto
import boto.s3.connection

endpoint   = 's3.epfl.ch'
access_key = 'put your access key here!'
secret_key = 'put your secret key here!'
bucket_id  = 'put your bucket id here!'

conn = boto.connect_s3(
    aws_access_key_id = access_key,
    aws_secret_access_key = secret_key,
    host = endpoint,
    calling_format = boto.s3.connection.OrdinaryCallingFormat(),
)

bucket = conn.get_bucket(bucket_id)

# listing objects in bucket
for key in bucket.list():
    print ("{name}\t{size}\t{modified}".format(
        name = key.name,
        size = key.size,
        modified = key.last_modified,
    ))

More information in the boto documentation.

chmod u+x s3.py
./s3.py
test    6   2021-06-23T08:29:48.409Z

rclone#

More information in the rclone documentation.

To configure, edit ~/.rclone.conf

[private]
type = s3
access_key_id = put_your_access_key_here
secret_access_key = put_your_secret_key_here
region = other-v2-signature
endpoint = https://s3.epfl.ch/

List a bucket content 
rclone ls private:<bucket_id>

Copy a file from your bucket 
rclone copy private:<bucket_id>/<PATH_TO_FILE> <DEST_PATH>
Ex: 
rclone copy private:32925-33bc8bbf6e9f71b235233ca33ce3b518/Doc/comics.pdf ~/Documents

Copy multiple files 
rclone copy --include "*.txt" private:<bucket_id>/<PATH_TO_FILES> <DEST_PATH>