s3m a command-line tool for storing streams of data in s3 buckets.
There are streams of data that can not be lost besides that when created, they degrade the performance of running systems, for example, if the stream is a backup of a database, every time the stream is produced it may lock the entire cluster (depends on the options and tools used mysqldump/xtrabackup for example), however, the effort to create the stream is proportional to the size of the database, in most cases the bigger the database is the more time and CPU and Memory are required.
In the case of backups, normally the streams are piped to a compression tool and later put in the disk, in some cases writing to the existing disk where the database is or to a remote mount endpoint, is not possible due to size constraints and the compressed backup should be streamed to am s3 bucket (X provider), therefore if for some reason the connection gets lost while streaming almost before finishing, the whole backup procedure could be corrupted and in worst scenario everything should start all over again.
The aim of s3m apart from trying to consume fewer resources is to make as much as possible "fault-tolerant" the storage procedure of the incoming stream so that even if the server lost network connectivity, the stream can be resumed and continue the upload process where it was left without the need to start all over again (no when input is from STDIN/pipe).
s3m is designed to run in very constrained environments with minimal resource usage:
- Disk Space: 512MB when streaming from STDIN (for buffering multipart uploads)
- Memory: Minimal - uses streaming to avoid loading data into RAM
- Network: Internet connection for S3 uploads (supports resumable uploads on disconnection)
Create ~/.config/s3m/config.yml:
hosts:
s3:
endpoint: s3.us-west-2.amazonaws.com
access_key: YOUR_ACCESS_KEY
secret_key: YOUR_SECRET_KEY
bucket: my-bucket # optional default buckethosts:
aws:
region: us-west-1
access_key: AWS_KEY
secret_key: AWS_SECRET
backblaze:
endpoint: s3.us-west-000.backblazeb2.com
access_key: B2_KEY
secret_key: B2_SECRETs3m uses the format: /host/bucket/file
# Upload file
s3m file.dat /s3/my-bucket/
# Upload with different name
s3m local-file.dat /s3/my-bucket/remote-name.dat
# Stream from STDIN
mariadb-dump database | s3m --pipe /s3/backups/db-backup.sqls3m get /s3/my-bucket/file.dat# List all buckets
s3m ls s3
# List objects in bucket
s3m ls s3/my-bucket
# List with prefix
s3m ls s3/my-bucket/path/# Bucket summary
s3m du s3/my-bucket
# Prefix summary grouped by UTC day
s3m du s3/my-bucket/backups/ --group-by day# List resumable multipart state
s3m streams
# Show one entry
s3m streams show <id>
# Clean broken/completed entries
s3m streams clean# One object uses DeleteObject
s3m rm s3/my-bucket/file.dat
# Multiple objects use DeleteObjects grouped by bucket
s3m rm s3/my-bucket/a.txt s3/my-bucket/b.txt
s3m rm s3/bucket-a/a.txt s3/bucket-b/b.txt# Delete all objects in the bucket, then delete the bucket
s3m rm -b --recursive s3/my-bucket# List objects as JSON
s3m ls s3/my-bucket --json
# Get object metadata as JSON
s3m get -m s3/my-bucket/file.dat --json
# Usage summary as JSON
s3m du s3/my-bucket --json
# Stream state as JSON
s3m streams ls --jsons3m cb s3/new-bucket# Delete object
s3m rm s3/my-bucket/file.dat
# Delete multiple objects
s3m rm s3/my-bucket/a.txt s3/my-bucket/b.txt
# Delete bucket
s3m rm -b s3/empty-bucket
# Recursively delete bucket contents, then the bucket
s3m rm -b --recursive s3/my-bucket# Compress before upload (uses Zstandard)
s3m --compress mysqldump.sql s3/backups/db.sql.zst# Generate secure encryption key (32 characters)
openssl rand -hex 16 > encryption.key
# Encrypt during upload
s3m --encrypt --enc-key "$(cat encryption.key)" file.dat s3/secure/file.dat.enc
# Decrypt during download
s3m get s3/secure/file.dat.enc --enc-key "$(cat encryption.key)"# Adjust part size for multipart uploads (in MB)
s3m --buffer 50 big-file.dat s3/large/huge-file.dat# Limit upload speed (in KB/s)
s3m --throttle 10240 file.dat s3/backups/file.dat # 10MB/s# Configure retry attempts for failed parts
s3m --retries 5 file.dat s3/bucket/file.dat- Regular file multipart uploads can be resumed.
STDIN/--pipeuploads are not resumable after interruption because the original input stream cannot be replayed safely.- When the input size is unknown,
s3muses a fixed multipart buffer of512 MiBper part. - For interrupted or failed multipart uploads from streaming paths, configure bucket lifecycle rules to clean up incomplete multipart uploads automatically.
# Unit + integration tests
cargo test
# Integration tests with MinIO (Podman)
just container # start MinIO (idempotent)
just test-integration # run MinIO-backed e2e tests against that MinIO
# or in one go
just container test-integration
# full suite (fmt + clippy + unit + integration)
just test
# stop/clean the container
podman rm -f s3m-test-minioExample config.yml for the default MinIO container:
---
hosts:
minio:
endpoint: http://localhost:9000
access_key: minioadmin
secret_key: minioadmincargo install cargo-llvm-cov
cargo llvm-cov --all-features --workspaceCoverage: 80%+
- Write tests for new features
- Run:
just test(to run the containerized integration tests) - Run:
cargo clippy --all-targets --all-features - Run:
cargo fmt