Buckets and objects

What a bucket, an object, an access key and the endpoint are, and how bucket names and addresses work.

Last updated 19 November 2026View as Markdown

Bucket

A named container for objects. Bucket names are unique across the whole platform, because each name is also a hostname: https://<bucket>.s3.<domain> is the bucket's Address. A name is 3–63 characters of lowercase letters, digits and hyphens, starting and ending with a letter or digit, without dots, two hyphens in a row or a leading xn-. Platform names and common words are reserved, and so are names starting with rds- (database backups) and kfk- (Kafka tiered storage); see Create a bucket. Once a bucket is deleted, its name is free again for anyone.

A bucket's name, zone and object lock are fixed when it is created. Its settings (versioning, lifecycle, replication, encryption, public access, website, CORS) each have a tab on the bucket's page.

Object

A file stored under a key, such as images/logo.png. Keys are case-sensitive and up to 1024 bytes. Folders are prefixes ending in /: the console lists a bucket by folder, as ListObjectsV2 with a / delimiter does.

Access key

An access key ID and a secret access key, which S3 clients sign requests with. An IAM access key works with S3 clients as well as with the API, the CLI and Terraform. A key acts as its IAM user or service account and can do what that account's policies allow. See Create an access key and Give access to one bucket.

Address

Every bucket has its own address, https://<bucket>.s3.<your platform domain>/<key>, whatever its zone: virtual-hosted style, the bucket is the first part of the host name. Clients that take an endpoint are given https://s3.<your platform domain> and add the bucket themselves with one setting. See S3 compatibility.

Object lock

A bucket created with object lock can lock object versions against deletion until a date (WORM), for backups and records you must keep. It turns versioning on for good and can only be chosen at creation.

Database backup bucket

Every managed database has a bucket named rds-<database id> that holds its encrypted backups. The bucket list shows its owner as Database backups. You can browse and download it; the platform writes it, and it is deleted together with the database. A Kafka cluster's tiered-storage bucket (kfk-…) works the same way and shows Kafka tiered storage as its owner.

Did this page help you?