Bcachefs Features¶

Compression¶

Bcachefs supports transparent compression of data at rest.

Enabling Compression¶

# At format time
bcachefs format --compression=zstd /dev/sdb

# As mount option
mount -t bcachefs -o compression=zstd /dev/sdb /mnt/data

Compression Algorithms¶

Algorithm	Speed	Ratio	Use Case
`lz4`	Very fast	Lower	General use, SSDs
`gzip`	Slow	Good	Archival
`zstd`	Fast	Better	Recommended default
`none`	N/A	None	Already compressed data

Background Compression¶

Set compression on an existing filesystem:

# Files written after this will be compressed
mount -o remount,compression=zstd /mnt/data

To compress existing data, copy files or use rebalance.

Encryption¶

Bcachefs provides whole-filesystem encryption using ChaCha20/Poly1305.

Creating an Encrypted Filesystem¶

# Interactive passphrase
bcachefs format --encrypted /dev/sdb

# With keyfile
bcachefs format --encrypted /dev/sdb
bcachefs unlock /dev/sdb < /path/to/keyfile

Unlocking for Mount¶

# Interactive unlock
bcachefs unlock /dev/sdb
mount -t bcachefs /dev/sdb /mnt/data

# With keyfile
bcachefs unlock /dev/sdb < /path/to/keyfile
mount -t bcachefs /dev/sdb /mnt/data

Key Management¶

# Add a new passphrase
bcachefs set-passphrase /dev/sdb

# Remove passphrase (must have another key)
bcachefs remove-passphrase /dev/sdb

Encryption vs LUKS¶

Aspect	Bcachefs Encryption	LUKS + Filesystem
Setup	Single step	Two layers
Performance	Native integration	Extra layer
Key management	Built-in	dm-crypt
Compatibility	Bcachefs only	Any filesystem

SSD Caching¶

Use fast devices to cache frequently accessed data from slower devices.

Tiered Storage Setup¶

bcachefs format \
    --label=hdd.hdd1 /dev/sda \
    --label=hdd.hdd2 /dev/sdb \
    --label=ssd.ssd1 /dev/nvme0n1p1 \
    --foreground_target=ssd \
    --promote_target=ssd \
    --background_target=hdd

Parameters:

Parameter	Description
`foreground_target`	Where new writes initially go
`promote_target`	Where hot data is promoted to
`background_target`	Where cold data is demoted to

Cache Behavior¶

New writes go to SSD (foreground)
Frequently read data promoted to SSD
Cold data migrated to HDD in background
Transparent to applications

Checksumming¶

All data and metadata is checksummed to detect corruption.

Checksum Algorithms¶

# At format time
bcachefs format --data_checksum=crc32c --metadata_checksum=crc64 /dev/sdb

Algorithm	Speed	Strength
`crc32c`	Fastest	Basic
`crc64`	Fast	Better
`xxhash`	Very fast	Good
`none`	N/A	No protection

Scrubbing¶

Verify all checksums and detect corruption:

bcachefs data scrub /mnt/data

With replication, corrupted data is automatically repaired from good copies.

Snapshots and Subvolumes¶

Subvolumes¶

Independent filesystem namespaces within a single bcachefs filesystem:

# Create subvolumes for different purposes
bcachefs subvolume create /mnt/data/documents
bcachefs subvolume create /mnt/data/photos
bcachefs subvolume create /mnt/data/backups

Benefits:

Independent snapshot policies
Can be mounted separately
Logical organization

Snapshots¶

Point-in-time copies sharing data blocks with the source:

# Create snapshot
bcachefs subvolume snapshot /mnt/data/documents /mnt/data/snapshots/documents-$(date +%Y%m%d)

# Read-only snapshot
bcachefs subvolume snapshot -r /mnt/data/documents /mnt/data/snapshots/documents-readonly

Snapshots are space-efficient - only changes consume additional space.

Quotas (Planned)¶

Quota support is planned but not yet fully implemented. Track progress in the bcachefs development roadmap.

Replication¶

Data Replication¶

# Store 2 copies of all data
bcachefs format --replicas=2 /dev/sdb /dev/sdc

Metadata Replication¶

Metadata can have higher replication than data:

bcachefs format \
    --replicas=1 \
    --metadata_replicas=2 \
    /dev/sdb /dev/sdc

Degraded Mode¶

Mount with missing devices:

mount -t bcachefs -o degraded /dev/sdb /mnt/data

Warning

Running degraded risks data loss if another device fails.

Performance Tuning¶

Discard/TRIM¶

Enable discard for SSDs:

mount -t bcachefs -o discard /dev/nvme0n1p1 /mnt/data

Noatime¶

Reduce write overhead by not updating access times:

mount -t bcachefs -o noatime /dev/sdb /mnt/data

Journal Location¶

Place journal on fast storage:

# Separate journal device (advanced)
bcachefs format --journal_dev=/dev/nvme0n1p2 /dev/sdb