For a list of features by their introduction, please see the table Changelog#By_feature.
The table below aims to serve as an overview for the stability status of the features BTRFS supports. While a feature may be functionally safe and reliable, it does not necessarily mean that its useful, for example in meeting your performance expectations for your specific workload. Combination of features can vary in performance, the table does not cover all possibilities.
The table is based on the latest released linux kernel: 4.17
The columns for each feature reflrect the status of the implementation in following ways:
Stability - completeness of the implementation, usecase coverage
Status since - kernel version when the status has been last changed
Performance - how much it could be improved until the inherent limits are hit
Notes - short description of the known issues, or other information related to status
- OK: should be safe to use, no known major defficiencies
- mostly OK: safe for general use, there are some known problems that do not affect majority of users
- Unstable: do not use for other then testing purposes, known severe problems, missing implementation of some core parts
|Trim (aka. discard)||OK||tbd||OK||fstrim and mounted with -o discard (has performance implications)|
|Defrag||mostly OK||tbd||OK||extents get unshared (see below)|
|Out-of-band dedupe||OK||tbd||mostly OK||(reflink), heavily referenced extents have a noticeable performance hit (see below)|
|File range cloning||OK||tbd||mostly OK||(reflink), heavily referenced extents have a noticeable performance hit (see below)|
|Auto-repair||OK||tbd||OK||automatically repair from a correct spare copy if possible (dup, raid1, raid10)|
|Scrub + RAID56||mostly OK||tbd||mostly OK|
|nodatacow||OK||tbd||OK||also see Manpage/btrfs(5).|
|Device replace||mostly OK||tbd||mostly OK||see below|
|Degraded mount||mostly OK||tbd||n/a||applies to raid levels with redundancy: needs at least two available devices always. Can get stuck in irreversible read-only mode if only one device is present.|
|Block group profile|
|Single (block group profile)||OK||tbd||OK|
|DUP (block group profile)||OK||tbd||OK|
|RAID1||OK||tbd||mostly OK||reading from mirrors in parallel can be optimized further (see below)|
|RAID10||OK||tbd||mostly OK||reading from mirrors in parallel can be optimized further (see below)|
|RAID56||Unstable||tbd||n/a||write hole still exists (see below)|
|Mixed block groups||OK||tbd||OK||see documentation|
|Filesystem resize||OK||tbd||OK||shrink, grow|
|Balance||OK||tbd||OK||balance + qgroups can be slow when there are many snapshots|
|Offline UUID change||OK||tbd||OK|
|Seeding||OK||tbd||OK||needs to be better documented|
|Quotas, qgroups||mostly OK||tbd||mostly OK||qgroups with many snapshots slows down balance|
|Free space tree||OK||4.9||OK||(see below)|
|no-holes||OK||tbd||OK||see documentation for compatibility|
|skinny-metadata||OK||tbd||OK||see documentation for compatibility|
|extended-refs||OK||tbd||OK||see documentation for compatibility|
Note to editors:
This page reflects status of the whole project and edits need to be approved by one of the maintainers (kdave). Suggest edits if:
- there's a known missing entry
- a particular feature combination that has a different status and is worth mentioning separately
- you knouw of a bug that lowers the feature status
- a reference could be enhanced by an actual link to documentation (wiki, manual pages)
Details that do not fit the table
The data affected by the defragmentation process will be newly written and will consume new space, the links to the original extents will not be kept. See also Manpage/btrfs-filesystem. Though autodefrag affects newly written data, it can read a few adjacent blocks (up to 64k) and write the contiguous extent to a new location. The adjacent blocks will be unshared. This happens on a smaller scale than the on-demand defrag and doesn't have the same impact.
Free space tree
- btrfs-progs support is read-only, ie. fsck can check the filesystem but is not able to keep the FST consistent and thus cannot run in repair mode
- the free space tree can be cleared using 'btrfs check --clear-space-cache v2' and will be rebuilt at next mount
Compatibility and historical references:
- btrfs-progs versions before v4.7.3 might accidentally do writes to the filesystem, but since there's no way to invalidate the FST, this causes inconsistency and possible corruption (using a piece of space twice). If you have made changes (btrfstune, repair, ...) to a FST enabled filesystem with btrfs progs, then mount with clear_cache,space_cache=v2 and hope the space written to was not reused yet. (see Status of free-space-tree feature)
- (fixed in linux 4.9) runtime support: fine on little-endian machines (x86*), known to be broken on big-endian (sparc64), see sparc64: btrfs module fails to load on big-endian machines
Out of band dedupe
File extents can be shared either due to snapshotting or reflink. As the number of owners of an extent grows, the time to process or modify the extent will also grow.
The deduplication increases the level of sharing and reduces data usage.
File range cloning
See the intro paragraph in Out of band dedupe. The range cloning increases the extent sharing.
The simple redundancy RAID levels utilize different mirrors in a way that does not achieve the maximum performance. The logic can be improved so the reads will spread over the mirrors evenly or based on device congestion.
Some fixes went to 4.12, namely scrub and auto-repair fixes. Feature marked as mostly OK for now.
Further fixes to raid56 related code are applied each release. The write hole is the last missing part, preliminary patches have been posted but needed to be reworked. The parity not checksummed note has been removed.
Device replace and device delete insist on being able to read or reconstruct all data. If any read fails due to an IO error, the delete/replace operation is aborted and the administrator must remove or replace the damaged data before trying again.
The filesystem disk format is stable. This means it is not expected to change unless there are very strong reasons to do so. If there is a format change, filesystems which implement the previous disk format will continue to be mountable and usable by newer kernels.
The core of the on-disk format that comprises building blocks of the filesystem:
- layout of the main data structures, eg. superblock, b-tree nodes, b-tree keys, block headers
- the COW mechanism, based on the original design of Ohad Rodeh's paper "Shadowing and clones"
Newly introduced features build on top of the above and could add specific structures. If a backward compatibility is not possible to maintain, a bit in the filesystem superblock denotes that and the level of incompatibility (full, read-only mount possible).