The Bcachefs project has just released version 1.33 as the “biggest new feature in the past ~2 years” for this modern copy‑on‑write Linux filesystem that supports encryption, snapshots, compression, and more, offering advanced features aimed at rivalling filesystems like Btrfs or ZFS.

The new version brings a major new “reconcile” engine that unifies data and metadata handling, automates replication and recovery, and substantially improves performance, logging, and error reporting under heavy load. But before we get to what’s new in this version, let’s quickly revisit the background.

As we informed you earlier, Bcachefs is undergoing a big transition in how it’s distributed and maintained upstream. In mid-2025, Linus Torvalds dropped Bcachefs from the official Linux kernel 6.17 merge window after a public dispute with lead developer Kent Overstreet.

With the release of Linux kernel 6.18 a week ago, the core Bcachefs code was removed entirely from the kernel tree. To keep Bcachefs alive and usable, Overstreet decided to distribute Bcachefs now as a DKMS module rather than part of the in-kernel filesystem lineup.

The new 1.33 version brings numerous improvements, with the most important being:

• A renamed and reworked “reconcile” mechanism (formerly “rebalance_v2”), now handling both user data and metadata, reacting to device or configuration changes, and automatically re-replicating degraded data or metadata without manual intervention.
• New indexing and scheduling for reconcile tasks, including high-priority handling of degraded extents and special treatment for rotational drives (HDDs), to avoid unnecessary I/O seeks.
• Several old commands — like data rereplicate or data job drop_extra_replicas — become obsolete. New commands such as reconcile status and reconcile wait replace them.
• Major backend improvements: asynchronous B-tree node merging, cleaned-up tracepoints and counters, better error logging (with clear differentiation between soft and hard errors), and enhanced progress reporting on recovery passes.
• New user-visible mount and filesystem options, such as mount_trusts_udev for device probing and writeback_timeout for controlling writeback timing — giving admins finer control.
• Work toward long-term codebase modernization in preparation for a future Rust migration — removal of many “goto”-style error-handling paths in favor of a cleaner, maintainable codebase structure.

For more information, see the announcement.

Bobby Borisov

Bobby, an editor-in-chief at Linuxiac, is a Linux professional with over 20 years of experience. With a strong focus on Linux and open-source software, he has worked as a Senior Linux System Administrator, Software Developer, and DevOps Engineer for small and large multinational companies.