3 .\" The DragonFly Project. All rights reserved.
5 .\" Redistribution and use in source and binary forms, with or without
6 .\" modification, are permitted provided that the following conditions
9 .\" 1. Redistributions of source code must retain the above copyright
10 .\" notice, this list of conditions and the following disclaimer.
11 .\" 2. Redistributions in binary form must reproduce the above copyright
12 .\" notice, this list of conditions and the following disclaimer in
13 .\" the documentation and/or other materials provided with the
15 .\" 3. Neither the name of The DragonFly Project nor the names of its
16 .\" contributors may be used to endorse or promote products derived
17 .\" from this software without specific, prior written permission.
19 .\" THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
20 .\" ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
21 .\" LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
22 .\" FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
23 .\" COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
24 .\" INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
25 .\" BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
26 .\" LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
27 .\" AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
28 .\" OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
29 .\" OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37 .Nd HAMMER file system
39 To compile this driver into the kernel,
40 place the following line in your
41 kernel configuration file:
42 .Bd -ragged -offset indent
46 Alternatively, to load the driver as a
47 module at boot time, place the following line in
49 .Bd -literal -offset indent
55 .Bd -literal -offset indent
56 /dev/ad0s1d[:/dev/ad1s1d:...] /mnt hammer rw 2 0
61 file system provides facilities to store file system data onto disk devices
62 and is intended to replace
64 as the default file system for
67 Among its features are instant crash recovery,
68 large file systems spanning multiple volumes,
69 data integrity checking,
71 fine grained history retention and snapshots,
72 pseudo-filesystems (PFSs),
73 mirroring capability and
74 unlimited number of files and links.
76 All functions related to managing
78 file systems are provided by the
88 For a more detailed introduction refer to the paper and slides listed in the
91 For some common usages of
100 .Ss Instant Crash Recovery
101 After a non-graceful system shutdown,
103 file systems will be brought back into a fully coherent state
104 when mounting the file system, usually within a few seconds.
108 .Ss Large File Systems & Multi Volume
111 file system can be up to 1 Exabyte in size.
112 It can span up to 256 volumes,
113 each volume occupies a
115 disk slice or partition, or another special file,
116 and can be up to 4096 TB in size.
119 file system size is 50 GB.
120 For volumes over 2 TB in size
124 normally need to be used.
134 .Ss Data Integrity Checking
136 has high focus on data integrity,
137 CRC checks are made for all major structures and data.
139 snapshots implements features to make data integrity checking easier:
140 The atime and mtime fields are locked to the ctime
141 for files accessed via a snapshot.
144 field is based on the PFS
146 and not on any real device.
147 This means that archiving the contents of a snapshot with e.g.\&
149 and piping it to something like
151 will yield a consistent result.
152 The consistency is also retained on mirroring targets.
153 .Ss Data Deduplication
154 To save disk space data deduplication can be used.
155 Data deduplication will identify data blocks which occur multiple times
156 and only store one copy, multiple reference will be made to this copy.
168 file system uses 64-bit transaction ids to refer to historical
169 file or directory data.
170 Transaction ids used by
172 are monotonically increasing over time.
174 when a transaction is made,
176 will always use higher transaction ids for following transactions.
177 A transaction id is given in hexadecimal format
180 .Li 0x00000001061a8ba6 .
191 .Ss History & Snapshots
192 History metadata on the media is written with every sync operation, so that
193 by default the resolution of a file's history is 30-60 seconds until the next
195 Prior versions of files and directories are generally accessible by appending
197 and a transaction id to the name.
198 The common way of accessing history, however, is by taking snapshots.
200 Snapshots are softlinks to prior versions of directories and their files.
201 Their data will be retained across prune operations for as long as the
203 Removing the softlink enables the file system to reclaim the space
204 again upon the next prune & reblock operations.
207 Version 3+ snapshots are also maintained as file system meta-data.
224 .Ss Pruning & Reblocking
225 Pruning is the act of deleting file system history.
226 By default only history used by the given snapshots
227 and history from after the latest snapshot will be retained.
228 By setting the per PFS parameter
230 history is guaranteed to be saved at least this time interval.
231 All other history is deleted.
232 Reblocking will reorder all elements and thus defragment the file system and
233 free space for reuse.
234 After pruning a file system must be reblocked to recover all available space.
235 Reblocking is needed even when using the
248 .Cm prune-everything ,
255 .Ss Pseudo-Filesystems (PFSs)
256 A pseudo-filesystem, PFS for short, is a sub file system in a
259 Each PFS has independent inode numbers.
262 file system is shared between all PFSs in it,
263 so each PFS is free to use all remaining space.
266 file system supports up to 65536 PFSs.
269 file system is PFS# 0, it is called the root PFS and is always a master PFS.
271 A PFS can be either master or slave.
272 Slaves are always read-only,
273 so they can't be updated by normal file operations, only by
275 operations like mirroring and pruning.
276 Upgrading slaves to masters and downgrading masters to slaves are supported.
278 It is recommended to use a
280 mount to access a PFS, except for root PFS;
281 this way no tools are confused by the PFS root being a symlink
282 and inodes not being unique across a
288 operations operates per PFS,
289 this includes mirroring, offline deduping, pruning, reblocking and rebalancing.
304 Mirroring is copying of all data in a file system, including snapshots
305 and other historical data.
306 In order to allow inode numbers to be duplicated on the slaves
308 mirroring feature uses PFSs.
309 A master or slave PFS can be mirrored to a slave PFS.
310 I.e.\& for mirroring multiple slaves per master are supported,
311 but multiple masters per slave are not.
319 .Cm mirror-read-stream ,
322 .Ss Fsync Flush Modes
325 file system implements several different
327 flush modes, the mode used is set via the
328 .Va vfs.hammer.flush_mode
332 .Ss Unlimited Number of Files and Links
333 There is no limit on the number of files or links in a
335 file system, apart from available disk space.
338 file systems support NFS export.
339 NFS export of PFSs is done using
341 mounts (for file/directory in root PFS
343 mount is not needed).
344 For example, to export the PFS
345 .Pa /hammer/pfs/data ,
350 and export the latter path.
352 Don't export a directory containing a PFS (e.g.\&
360 above) should be exported (subdirectory may be escaped if exported).
361 .Ss File System Versions
362 As new features have been introduced to
364 a version number has been bumped.
367 file system has a version, which can be upgraded to support new features.
373 .Cm version-upgrade ;
377 .Ss Preparing the File System
378 To create and mount a
387 file systems must have a unique name on a per-machine basis.
388 .Bd -literal -offset indent
389 newfs_hammer -L HOME /dev/ad0s1d
390 mount_hammer /dev/ad0s1d /home
393 Similarly, multi volume file systems can be created and mounted by
394 specifying additional arguments.
395 .Bd -literal -offset indent
396 newfs_hammer -L MULTIHOME /dev/ad0s1d /dev/ad1s1d
397 mount_hammer /dev/ad0s1d /dev/ad1s1d /home
400 Once created and mounted,
402 file systems need periodic clean up making snapshots, pruning and reblocking,
403 in order to have access to history and file system not to fill up.
404 For this it is recommended to use the
412 .Nm hammer Cm cleanup
416 It is also possible to perform these operations individually via
418 For example, to reblock the
420 file system every night at 2:15 for up to 5 minutes:
421 .Bd -literal -offset indent
422 15 2 * * * hammer -c /var/run/HOME.reblock -t 300 reblock /home \e
430 command provides several ways of taking snapshots.
431 They all assume a directory where snapshots are kept.
432 .Bd -literal -offset indent
434 hammer snapshot /home /snaps/snap1
435 (...after some changes in /home...)
436 hammer snapshot /home /snaps/snap2
441 point to the state of the
443 directory at the time each snapshot was taken, and could now be used to copy
444 the data somewhere else for backup purposes.
448 is set up to create nightly snapshots of all
452 and to keep them for 60 days.
454 A snapshot directory is also the argument to the
457 command which frees historical data from the file system that is not
458 pointed to by any snapshot link and is not from after the latest snapshot
461 .Bd -literal -offset indent
466 Mirroring is set up using
468 pseudo-filesystems (PFSs).
469 To associate the slave with the master its shared UUID should be set to
470 the master's shared UUID as output by the
471 .Nm hammer Cm pfs-master
473 .Bd -literal -offset indent
474 hammer pfs-master /home/pfs/master
475 hammer pfs-slave /home/pfs/slave shared-uuid=<master's shared uuid>
480 link is unusable for as long as no mirroring operation has taken place.
482 To mirror the master's data, either pipe a
486 or, as a short-cut, use the
488 command (which works across a
491 Initial mirroring operation has to be done to the PFS path (as
493 can't access it yet).
494 .Bd -literal -offset indent
495 hammer mirror-copy /home/pfs/master /home/pfs/slave
498 It is also possible to have the target PFS auto created
499 by just issuing the same
501 command, if the target PFS doesn't exist you will be prompted
502 if you would like to create it.
503 You can even omit the prompting by using the
506 .Bd -literal -offset indent
507 hammer -y mirror-copy /home/pfs/master /home/pfs/slave
510 After this initial step
512 mount can be setup for
513 .Pa /home/pfs/slave .
514 Further operations can use
517 .Bd -literal -offset indent
518 mount_null /home/pfs/master /home/master
519 mount_null /home/pfs/slave /home/slave
521 hammer mirror-copy /home/master /home/slave
524 To NFS export from the
530 without PFSs, and the PFS
531 .Pa /hammer/pfs/data ,
541 .Bd -literal -offset indent
542 /hammer/pfs/data /hammer/data null rw
549 .Bd -literal -offset indent
555 .It "hammer: System has insuffient buffers to rebalance the tree. nbuf < %d"
558 PFS uses quite a bit of memory and
559 can't be done on low memory systems.
560 It has been reported to fail on 512MB systems.
561 Rebalancing isn't critical for
563 file system operation;
590 .%O http://www.dragonflybsd.org/hammer/hammer.pdf
591 .%T "The HAMMER Filesystem"
596 .%O http://www.dragonflybsd.org/hammer/nycbsdcon/
597 .%T "Slideshow from NYCBSDCon 2008"
602 .%O http://www.ntecs.de/sysarch09/HAMMER.pdf
603 .%T "Slideshow for a presentation held at KIT (http://www.kit.edu)"
605 .Sh FILESYSTEM PERFORMANCE
608 file system has a front-end which processes VNOPS and issues necessary
609 block reads from disk, and a back-end which handles meta-data updates
610 on-media and performs all meta-data write operations.
611 Bulk file write operations are handled by the front-end.
614 defers meta-data updates virtually no meta-data read operations will be
615 issued by the frontend while writing large amounts of data to the file system
616 or even when creating new files or directories, and even though the
617 kernel prioritizes reads over writes the fact that writes are cached by
618 the drive itself tends to lead to excessive priority given to writes.
620 There are four bioq sysctls, shown below with default values,
621 which can be adjusted to give reads a higher priority:
622 .Bd -literal -offset indent
623 kern.bioq_reorder_minor_bytes: 262144
624 kern.bioq_reorder_burst_bytes: 3000000
625 kern.bioq_reorder_minor_interval: 5
626 kern.bioq_reorder_burst_interval: 60
629 If a higher read priority is desired it is recommended that the
630 .Va kern.bioq_reorder_minor_interval
631 be increased to 15, 30, or even 60, and the
632 .Va kern.bioq_reorder_burst_bytes
633 be decreased to 262144 or 524288.
637 file system first appeared in
643 file system was designed and implemented by
644 .An Matthew Dillon Aq dillon@backplane.com ,
645 data deduplication was added by
647 This manual page was written by
650 .An Thomas Nikolajsen .
652 Data deduplication is considered experimental.