#tar-archive #amazon-s3 #s3 #archive #tar #object-storage #helper


Library of testing helpers to make it easier to write test fixtures. Not for use outside of ssstar crate. Breaking changes may ocurr without warning.

17 unstable releases (6 breaking)

0.7.3 Mar 12, 2024
0.7.1 Nov 20, 2023
0.6.0 Jul 19, 2023
0.4.3 Mar 3, 2023
0.2.0 Sep 2, 2022

#888 in Filesystem

Download history 29/week @ 2024-03-14 70/week @ 2024-03-21 14/week @ 2024-03-28 4/week @ 2024-04-04

470 downloads per month
Used in ssstar

Apache-2.0 OR MIT

572 lines


ssstar is a Rust library crate as well as a command-line tool to create and extract tar-compatible archives containing objects stored in S3 or S3-compatible storage. It works similarly to GNU tar, and produces archives that are 100% compatible with tar, though it uses different command line arguments.

Crates.io Docs.rs CI Coverage Status Crates.io

ssstar provides a cross-platform Rust-powered CLI as well as a Rust library crate that lets you create tar archives containing objects from S3 and S3-compatible object storage, regardless of size. ssstar applies concurrency aggressively, and uses a streaming design which means even multi-TB objects can be processed with minimal memory utilization. The resulting tar archive can itself be uploaded to object storage, written to a local file, or written to stdout and piped to another command line tool.

We built ssstar so our customers using the elastio cloud native backup and recovery CLI could backup and restore S3 buckets directly into and from Elastio vaults, however we made the tool generic enough that it can be used by itself whenever you need to package one or more S3 objects into a tarball.



On any supported platform (meaning Windows, macOS (both Intel and Apple Silicon) and Linux), if you have a recent Rust compiler installed you can use cargo install to get the ssstar CLI:

  • Ensure you have at least Rust 1.63.0 installed by following this guide.
  • Run cargo install ssstar-cli --locked to compile ssstar from source and install locally.

Precompiled binaries

See the GitHub Releases for pre-compiled binaries for Windows, mac, and Linux.

Usage (without Elastio)

To create a tar archive, you specify S3 buckets, objects, entire prefixes, or globs, as well as where you want the tar archive to be written:

# Archive an entire bucket and write the tar archive to another bucket
ssstar create \
  s3://my-source-bucket \
  --s3 s3://my-destination-bucket/backup.tar

# Archive all objects in the `foo/` prefix (non-recursive) and write the tar archive to a local file
ssstar create \
  s3://my-source-bucket/foo/ \
  --file ./backup.tar

# Archive some specific objects identified by name, and write the tar archive to stdout and pipe that to
# gzip
ssstar create \
  s3://my-source-bucket/object1 s3://my-source-bucket/object2 \
  --stdout | gzip > backup.tar.gz

# Archive all objects matching a glob, and write the tar archive to another bucket
ssstar create \
  "s3://my-source-bucket/foo/**" \
  --s3 s3://my-destination-bucket/backup.tar

You can pass multiple inputs to ssstar create, using a mix of entire buckets, prefixes, specific objects, and globs. Just make sure that when you use globs you wrap them in quotes, otherwise your shell may try to evaluate them. For example:

# Archive a bunch of different inputs, writing the result to a file
ssstar create \
  s3://my-source-bucket/                  \ # <-- include all objects in `my-source-bucket`
  s3://my-other-bucket/foo/               \ # <-- include all objects in `foo/` (non-recursive)
  s3://my-other-bucket/bar/boo            \ # <-- include the object with key `bar/boo`
  "s3://yet-another-bucket/logs/2022*/**" \ # <-- recursively include all objects in any prefix `logs/2022*`
  --file ./backup.tar                     # <-- this is the path where the tar archive will be written

To extract a tar archive and write the contents directly to S3 objects, you specify where to find the tar archive, optional filters to filter what is extracted, and the S3 bucket and prefix to which to extract the contents.

A simple example:

# Extract a local tar archive to the root of an S3 bucket `my-bucket`
ssstar extract --file ./backup.tar s3://my-bucket

Each file in the tar archive will be written to the bucket my-bucket, with the object key equal to the file path within the archive. For example if the archive contains a file foo/bar/baz.txt, that file will be written to s3://my-bucket/foo/bar/baz.txt.

You can provide not just a target bucket but also a prefix as well, e.g.:

# Extract a local tar archive to the prefix `restored/` of an S3 bucket `my-bucket`
ssstar extract --file ./backup.tar s3://my-bucket/restored/

In that case, if the tar archive contains a file foo/bar/baz.txt, it will be written to s3://my-bucket/restored/foo/bar/baz.txt. NOTE: In S3, prefixes don't necessarily end in /; if you don't provide the trailing / character to the S3 URL passed to ssstar extract, it will not be added for you! Instead you'll get something like s3://my-bucket/restoredfoo/bar/baz, which may or may not be what you actually want!

If you don't want to extract the full contents of the archive, you can specify one or more filters. These can be exact file paths, directory paths ending in /, or globs. For example:

ssstar extract --file ./backup.tar \
  foo/bar/baz.txt          \ # <-- extract the file `foo/bar/baz.txt` if it's present in the archive
  boo/                     \ # <-- extract all files in the `boo` directory (recursive)
  "baz/**/*.txt"           \ # <-- extract any `.txt` file anywhere in `baz/`, recursively
  s3://my-bucket/restored/   # <-- write all matching files to the `restored/` prefix in `my-bucket`

Usage (with Elastio)

To use with Elastio, create archives with the --stdout option and pipe to elastio stream backup, and restore them by piping elastio stream restore to ssstar extract with the --stdin option. For example:

# Backup an entire S3 bucket `my-source-bucket` to the default Elastio vault:
ssstar create s3://my-source-bucket/ --stdout \
  | elastio stream backup --hostname-override my-source-bucket --stream-name my-backup
# Restore a recovery point with ID `$RP_ID` from Elastio to the `my-destination-bucket` bucket:
elastio stream restore --rp $RP_ID \
  | ssstar extract --stdin s3://my-destination-bucket

For more about using the Elastio CLI, see the Elastio CLI docs

Advanced CLI Options

Run ssstar create --help and ssstar extract --help to get the complete CLI usage documentation for archive creation and extraction, respectively. There are a few command line options that are particularly likely to be of interest:

Using a custom S3 endpoint

ssstar is developed and tested against AWS S3, however it should work with any object storage system that provides an S3-compatible API. In particular, most of the automated tests our CI system runs actually use Minio and not the real S3 API. To use ssstar with an S3-compatible API, use the --s3-endpoint option. For example, if you have a Minio server running at, using default minioadmin credentials, you can use it with ssstar like this:

ssstar --s3-endpoint \
  --aws-access-key-id minioadmin --aws-secret-access-key minio-admin \

Controlling Concurrency

The --max-concurrent-requests argument controls how many concurrent S3 API operations will be performed in each stage of the archive creation or extraction process. The default is 10, because that is what the AWS CLI uses. However if you are running ssstar on an EC2 instance with multi-gigabit Ethernet connectivity to S3, 10 concurrent requests may not be enough to saturate the network connection. Experiment with larger values to see if you experience faster transfer times with more concurrency.

Usage (in a Rust project)

The library crate ssstar is the engine that powers the ssstar CLI. When we wrote ssstar we deliberately kept all of the functionality in a library crate with a thin CLI wrapper on top, because ssstar is being used internally in Elastio to power our upcoming S3 backup feature. You too can integrate ssstar functionality into your Rust application. Just add ssstar as a dependency in your Cargo.toml:

ssstar = "0.7.3"

See the docs.rs documentation for ssstar for more details and some examples. You can also look at the ssstar CLI code ssstar-cli/main.rs to see how we implemented our CLI in terms of the ssstar library crate.


Licensed under either of

at your option.


Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.



~753K SLoC