The complete release log for MooseFS — from 4.0.0 (February 2017, which introduced erasure coding) through the current 5.0.x stream. Entries are listed reverse-chronologically and grouped by major version. Each release lists every change verbatim, prefixed with the component it touches.
Looking for the high-level story instead? See the history page. For supported version windows and end-of-life dates, see versions.
gui added exit after CS section when there are no active masters
master when chunk has both bits ARCH and TRASH set then system will use TRASH class (if defined) first and then ARCH (if defined)
4.36.4
cs added support for new .chunkdb format to mfschunkdbdump (intr. in 4.36.3)
master fixed reporting partial ec as wrong version
freebsd added mfschunkdbdump with manpage to plist file
4.36.3
cs added ‘shared’ mode for each hdd with optional available space limit (size or percent)
master wait with next replication when previous returned status ‘time out’
cs changed calculation of estimated disk usage (using real disk usage from stat instead of file length)
cs added ‘percent’ limit modes for hard drives
4.36.2
master fixed handling writecounter on failed write operations
gui fixed compatibility with MFS 3.0 issue
cs added timeout to replicator (const 150 seconds) with progress check
all changed trashtime to trashretention in code, manpages etc.
4.36.1
cs fixed NOP handling during write to chunk
mount increased waiting time for write finish
cs added option to switch between mmap and malloc (default set to malloc)
client removed using mmap for master buffer allocations
4.36.0
master+tools fixed quick version of mfsfileinfo (support for EC4)
man updated man pages
master+cs added error message when a config file can’t be loaded
client fixed invalidating directory cache after unlink, rmdir and rename
cs fixed cond variables leaking
4.35.1
libpro updated API for version 4.35
mount fixed printing version on MacOS
4.35.0
master+cli+gui added info about maintenance mode with timeout (how many seconds left)
cs added stalled jobs detection
cs added more jobs info to report (avg job time, max job time etc.)
master reduced number of copies that will be cleaned up on each call after chunkserver disconnection
4.34.2
master fixed state switching (made DUMMY the final state — PRO only)
cli added chart data type ‘all’ — an alias to all charts (master and chunkserver)
cli when in chunkserver data host and port are not specified then all available chunkservers will be used
4.34.1
master added job exit reasons to info file
master fixed failed jobs algorithm (rebalance stalled when there was low rebalance probability)
4.34.0
master made chunks always readable after ‘mfsfilerepairing’ them (will return zeros when chunks are missing)
master fixed logical error (master without metadata could become ELECT and then LEADER which led to segfault)
all new feature: minimum file length needed for being moved from KEEP to ARCH mode in storage class
4.33.2
cs+master removed register protocols compatible with pre 2.0 versions
4.33.1
cs fixed crc data in info packets (*_CHUNK_CHECKSUM, *_CHUNK_CHECKSUM_TAB, *_CHUNK_INFO)
mount added NOP packets to master proxy (fixed query error on long requests from tools)
metasearch fixed parsing hex numbers with letters
master added changelog after automatic version fix in chunk_do_jobs
master added doing extra chunk job when chunk is removed from delay queue
4.33.0
cs+gui added clearing last hdd errors in chunkservers
cli added JSON output (pull request by Aleksander Wieliczko)
4.32.1
gui fixed using & in javascript URLs
cs added tracking of busy chunks (don’t send chunk status when chunk is busy)
master removed tracking of busy chunks
master added chunk deletion delay after replication (mainly during EC<->COPY conversion)
master fixed emergency save in FOLLOWER
gui added tooltips to charts
master always return MFS_ERROR_EAGAIN when number of valid copies is zero but goal equivalent is higher than zero
master+cs fixed overflows in old charts (only for 3.x gui compatibility)
cgiserv fixed error/redirect responses (pull request for 3.x by @Eronana)
4.32.0
metatools new tool mfsmetasearch
cs+master added to load packet info about disk scanning and removals
master never use registering chunkserver as a destination place for replication (potential race between replication and chunk registration)
master do not check files when system is during chunkserver disconnections, registration etc.
cs reduce number of chunks per packet sent by chunkserver during registration (avoid master clog)
master kill meta store process after becoming LEADER
master+cs prepare system for integrity testing between master and chunkserver
master fixed handling copy sgid for directories on FOLLOWER and during restore from changelog
master added new charts data (total data I/O from all clients) [sic]
master+supervisor added option to force metadata store (PRO only feature)
charts fixed leap years calculation for years that are multiples of 400
master fixed changelog order between unlock and close file or session
master removing good chunks from list of missing chunks sent to cli/gui
master+cs changed packet for access to nonexistent chunk (from “chunk lost” to “chunk doesn’t exist” — prevents turning off replications)
client added limit for sustained parents that can be stored
master fixed handling license quotas (PRO only feature)
4.31.9
master fixed changelog rotation (intr. in 4.31.1)
4.31.8
master take into account parts on the same servers in redundancy level calculations (for example — do not remove copies when all parts are still on the same server after split)
cs force fsync and close after replication
4.31.7
master fixed redundancy level calculating algorithm (intr. in 4.31.5)
mount fixed handling invalid arguments in posix lock command
4.0.7
mount+master added update master version during reconnection
master fixed winattr field size in setattr command
mount fixed dirattrcache (freezes or inefficiencies during listing directories)
gui fixed displaying master charts without LEADER (PRO only)
mount added info about using open dir cache in .oplog/.ophistory
4.0.6
master removed starting protection time from client communication module
4.0.5
mount fixed condition in assertion (read module)
4.0.4
mount fixed timeout syslog message during read data
mount timed out worker that was able to read anything doesn’t increase try counter
mount moved retries counter from file record to request record
4.0.3
mount fixed read from split chunk
mount fixed support for CREATE/REPLACE flags in setxattr
mount added headers for flock defines
all added check for poll.h header file
master added test for WRITE access on directory during moving between different parents
master added clearing suid/sgid during write
mount increased performance when reading from chunk divided into parts (EC support)
4.0.2
mount fixed reading from chunk divided into parts (EC support)
4.0.1
all removed PRO/CE bit from version number
4.0.0
all added erasure codes
tools added new commands mfsXXXtrashretention (deprecate mfsXXXtrashtime commands)
tools added new commands mfsXXXfacl for managing ACLs on non-Linux OS’es
master added new storage mode (trash) in storage classes
Next steps
Ready to upgrade?
If you’re moving between major versions, the documentation has the step-by-step upgrade guide. To see which streams are still actively supported, head to the versions page.