The object store has moved

You have reached a page that used to be the MinIO console.

The endpoint

S3 endpoint
https://s3.smtornado.com
signature v4, path-style
Region
lab
the configured Garage region

⚠ Use path-style addressing (https://s3.smtornado.com/bucket/key), not virtual-host style (https://bucket.s3.smtornado.com/key). Some clients default to virtual-host and fail with a DNS error that looks like the endpoint is down.

Use your newly issued Garage access key and secret. Old MinIO keys do not work. Each credential has its own reviewed bucket permissions.

MinIO client (mc)

mc alias set lab https://s3.smtornado.com ACCESS_KEY SECRET_KEY --api S3v4

mc ls lab/
mc ls lab/datasets
mc cp lab/datasets/table.csv .
mc mirror lab/datasets ./datasets     # a whole prefix

The same mc you used with MinIO. Only the alias changes. ⚠ mc admin commands do not work — those are MinIO server administration and Garage is a different server.

AWS CLI

aws configure --profile lab
#   Access Key ID     : ACCESS_KEY
#   Secret Access Key : SECRET_KEY
#   Default region    : lab
#   Default output    : json

aws --profile lab --endpoint-url https://s3.smtornado.com s3 ls
aws --profile lab --endpoint-url https://s3.smtornado.com s3 ls s3://datasets/
aws --profile lab --endpoint-url https://s3.smtornado.com \
    s3 cp s3://datasets/table.csv .

⚠ --endpoint-url on every command, or the CLI talks to Amazon. To avoid repeating it, put endpoint_url = https://s3.smtornado.com under [profile lab] in ~/.aws/config — supported by recent AWS CLI v2.

Python (boto3)

import boto3
from botocore.config import Config

s3 = boto3.client(
    "s3",
    endpoint_url="https://s3.smtornado.com",
    aws_access_key_id="ACCESS_KEY",
    aws_secret_access_key="SECRET_KEY",
    region_name="lab",
    # path-style addressing — see the note above
    config=Config(s3={"addressing_style": "path"},
                  signature_version="s3v4"),
)

for page in s3.get_paginator("list_objects_v2").paginate(
        Bucket="datasets", Prefix=""):
    for obj in page.get("Contents", []):
        print(obj["Key"], obj["Size"])

s3.download_file("datasets", "table.csv", "table.csv")

Keep the credentials out of the source file — boto3 reads ~/.aws/credentials, or the environment (AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY), if you drop the two arguments.

⚠ What changed, and what broke

Your bucket names and stored objects unchanged
Authenticated S3 API — mc, aws, boto3, rclone works
Multipart upload, ListObjectsV2 works
The web console gone — Garage has none
Anonymous (unsigned) S3 API access gone
Bucket policies and ACLs not implemented — permissions are per access key
Object versioning / object lock not implemented — recovery snapshots do not provide S3 versioning or Object Lock

⚠ If you have code doing unsigned S3 calls — boto3 with UNSIGNED, or a bucket you made public with mc anonymous — it will fail against the new store, and no change of URL fixes it. Garage implements neither bucket policies nor ACLs; permission is per access key. Either use a key, or use the public HTTPS links below, which serve the same objects over ordinary anonymous HTTP.

Datasets that were published with a public link still resolve at the URLs they were published at, anonymously, in a browser and with curl and wget. Those links were not changed and will not be.

Uploading

mc cp ./big.fits lab/datasets/
aws --profile lab --endpoint-url https://s3.smtornado.com \
    s3 cp ./big.fits s3://datasets/

⚠ Large uploads go straight to the origin rather than through the CDN, so a single request is not capped at 100 MB. If you see 413 on an upload, your client is not using this endpoint — check for a proxy or an old alias.