1 .\" Copyright (c) 2007 The DragonFly Project. All rights reserved.
3 .\" This code is derived from software contributed to The DragonFly Project
4 .\" by Matthew Dillon <dillon@backplane.com>
6 .\" Redistribution and use in source and binary forms, with or without
7 .\" modification, are permitted provided that the following conditions
10 .\" 1. Redistributions of source code must retain the above copyright
11 .\" notice, this list of conditions and the following disclaimer.
12 .\" 2. Redistributions in binary form must reproduce the above copyright
13 .\" notice, this list of conditions and the following disclaimer in
14 .\" the documentation and/or other materials provided with the
16 .\" 3. Neither the name of The DragonFly Project nor the names of its
17 .\" contributors may be used to endorse or promote products derived
18 .\" from this software without specific, prior written permission.
20 .\" THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21 .\" ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22 .\" LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23 .\" FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
24 .\" COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25 .\" INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
26 .\" BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27 .\" LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
28 .\" AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
29 .\" OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
30 .\" OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 .\" $DragonFly: src/sbin/hammer/hammer.8,v 1.58 2008/11/13 02:04:27 dillon Exp $
40 .Nd HAMMER file system utility
49 .\" .Op Fl s Ar linkpath
53 .Op Fl C Ar cachesize Ns Op Ns Cm \&: Ns Ar readahead
58 This manual page documents the
60 utility which provides miscellaneous functions related to managing a
63 For a general introduction to the
65 file system, its features, and
66 examples on how to set up and maintain one, see
69 The options are as follows:
70 .Bl -tag -width indent
74 Tell the mirror commands to use a 2-way protocol, which allows
75 automatic negotiation of transaction id ranges.
76 This option is automatically enabled by the
80 Specify a bandwidth limit in bytes per second for mirroring streams.
81 This option is typically used to prevent batch mirroring operations from
82 loading down the machine.
83 The bandwidth may be suffixed with
87 to specify values in kilobytes, megabytes, and gigabytes per second.
88 If no suffix is specified, bytes per second is assumed.
90 Unfortunately this is only applicable to the pre-compression bandwidth
91 when compression is used, so a better solution would probably be to
98 When pruning, rebalancing or reblocking you can tell the utility
99 to start at the object id stored in the specified file.
100 If the file does not exist
102 will start at the beginning.
106 specific period of time and is unable to complete the operation it will
107 write out the current object id so the next run can pick up where it left off.
110 runs to completion it will delete
113 Specify the volumes making up a
117 is a colon-separated list of devices, each specifying a
121 When maintaining a streaming mirroring this option specifies the
122 minimum delay after a batch ends before the next batch is allowed
124 The default is five seconds.
131 specification for the source and/or destination.
133 Decrease verboseness.
134 May be specified multiple times.
136 Specify recursion for those commands which support it.
138 When pruning, rebalancing or reblocking you can tell the utility to stop
139 after a certain period of time.
140 This option is used along with the
142 option to prune, rebalance or reblock incrementally.
144 Increase verboseness.
145 May be specified multiple times.
147 Force "yes" for any interactive question.
151 will not attempt to break-up large initial bulk transfers into smaller
153 This can save time but if the link is lost in the middle of the
154 initial bulk transfer you will have to start over from scratch.
155 This option is not recommended.
156 For more information see the
159 .It Fl C Ar cachesize Ns Op Ns Cm \&: Ns Ar readahead
160 Set the memory cache size for any raw
167 for megabytes is allowed,
168 else the cache size is specified in bytes.
170 The read-behind/read-ahead defaults to 4
174 This option is typically only used with diagnostic commands
175 as kernel-supported commands will use the kernel's buffer cache.
176 .It Fl S Ar splitsize
177 Specify the bulk splitup size in bytes for mirroring streams.
182 will do an initial run-through of the data to calculate good
183 transaction ids to cut up the bulk transfers, creating
184 restart points in case the stream is interrupted.
185 If we don't do this and the stream is interrupted it might
186 have to start all over again.
187 The default is a splitsize of 4G.
189 At the moment the run-through is disk-bandwidth-heavy but some
190 future version will limit the run-through to just the B-Tree
191 records and not the record data.
193 The splitsize may be suffixed with
197 to specify values in kilobytes, megabytes, or gigabytes.
198 If no suffix is specified, bytes is assumed.
200 When mirroring very large filesystems the minimum recommended
202 A small split size may wind up generating a great deal of overhead
203 but very little actual incremental data and is not recommended.
205 Enable compression for any remote ssh specifications.
208 option has already been reserved for other purposes so we had to use
210 This option is typically used with the mirroring directives.
212 Force "yes" for any interactive question.
215 The commands are as follows:
216 .Bl -tag -width indent
217 .\" ==== synctid ====
218 .It Cm synctid Ar filesystem Op Cm quick
219 Generates a guaranteed, formal 64 bit transaction id representing the
220 current state of the specified
223 The file system will be synced to the media.
227 keyword is specified the file system will be soft-synced, meaning that a
228 crash might still undo the state of the file system as of the transaction
229 id returned but any new modifications will occur after the returned
230 transaction id as expected.
232 This operation does not create a snapshot.
233 It is meant to be used
234 to track temporary fine-grained changes to a subset of files and
235 will only remain valid for
237 snapshot access purposes for the
239 period configured for the PFS.
240 If you desire a real snapshot then the
242 directive may be what you are looking for.
244 .It Cm bstats Op Ar interval
247 B-Tree statistics until interrupted.
250 seconds between each display.
251 The default interval is one second.
252 .\" ==== iostats ====
253 .It Cm iostats Op Ar interval
257 statistics until interrupted.
260 seconds between each display.
261 The default interval is one second.
262 .\" ==== history ====
263 .It Cm history Ar path ...
264 Show the modification history for
266 file's inode and data.
267 .\" ==== blockmap ====
269 Dump the blockmap for the file system.
272 blockmap is two-layer
273 blockmap representing the maximum possible file system size of 1 Exabyte.
274 Needless to say the second layer is only present for blocks which exist.
276 blockmap represents 8-Megabyte blocks, called big-blocks.
277 Each big-block has an append
278 point, a free byte count, and a typed zone id which allows content to be
279 reverse engineered to some degree.
283 allocations are essentially appended to a selected big-block using
284 the append offset and deducted from the free byte count.
285 When space is freed the free byte count is adjusted but
287 does not track holes in big-blocks for reallocation.
288 A big-block must be completely freed, either
289 through normal file system operations or through reblocking, before
292 Data blocks can be shared by deducting the space used from the free byte
293 count for each shared references.
294 This means the free byte count can legally go negative.
296 This command needs the
299 .\" ==== checkmap ====
301 Check the blockmap allocation count.
303 will scan the B-Tree, collect allocation information, and
304 construct a blockmap in-memory. It will then check that blockmap
305 against the on-disk blockmap.
307 This command needs the
311 .It Cm show Op Ar lo Ns Cm \&: Ns Ar objid
313 By default this command will validate all B-Tree
314 linkages and CRCs, including data CRCs, and will report the most verbose
315 information it can dig up.
316 Any errors will show up with a
318 in column 1 along with various
321 If you specify a localization field or a localization:obj_id field,
322 .Ar lo Ns Cm \&: Ns Ar objid ,
324 search for the key printing nodes as it recurses down, and then
325 will iterate forwards.
326 These fields are specified in HEX.
327 Note that the pfsid is the top 16 bits of the 32 bit localization
328 field so PFS #1 would be 00010000.
332 the command will report less information about the inode contents.
336 the command will not report the content of the inode or other typed
341 the command will not report volume header information, big-block fill
342 ratios, mirror transaction ids, or report or check data CRCs.
343 B-Tree CRCs and linkages are still checked.
345 This command needs the
348 .\" ==== show-undo ====
354 This command needs the
358 .\" Dump the B-Tree, record, large-data, and small-data blockmaps, showing
359 .\" physical block assignments and free space percentages.
360 .\" ==== recover ====
361 .It Cm recover Ar targetdir
362 This is a low level command which operates on the filesystem image and
363 attempts to locate and recover files from a corrupted filesystem. The
364 entire image is scanned linearly looking for B-Tree nodes. Any node
365 found which passes its crc test is scanned for file, inode, and directory
366 fragments and the target directory is populated with the resulting data.
367 files and directories in the target directory are initially named after
368 the object id and are renamed as fragmentory information is processed.
370 This command keeps track of filename/objid translations and may eat a
371 considerably amount of memory while operating.
373 This command is literally the last line of defense when it comes to
374 recovering data from a dead filesystem.
375 .\" ==== namekey1 ====
376 .It Cm namekey1 Ar filename
379 64 bit directory hash for the specified file name, using
380 the original directory hash algorithm in version 1 of the file system.
381 The low 32 bits are used as an iterator for hash collisions and will be
383 .\" ==== namekey2 ====
384 .It Cm namekey2 Ar filename
387 64 bit directory hash for the specified file name, using
388 the new directory hash algorithm in version 2 of the file system.
389 The low 32 bits are still used as an iterator but will start out containing
390 part of the hash key.
391 .\" ==== namekey32 ====
392 .It Cm namekey32 Ar filename
393 Generate the top 32 bits of a
395 64 bit directory hash for the specified file name.
398 Shows extended information about all the mounted
401 The information is divided into sections:
402 .Bl -tag -width indent
403 .It Volume identification
404 General information, like the label of the
406 filesystem, the number of volumes it contains, the FSID, and the
409 .It Big block information
410 Big block statistics, such as total, used, reserved and free big blocks.
411 .It Space information
412 Information about space used on the filesystem.
413 Currently total size, used, reserved and free space are displayed.
415 Basic information about the PFSs currently present on a
420 is the ID of the PFS, with 0 being the root PFS.
422 is the current snapshot count on the PFS.
424 displays the mount point of the PFS is currently mounted on (if any).
426 .\" ==== cleanup ====
427 .It Cm cleanup Op Ar filesystem ...
428 This is a meta-command which executes snapshot, prune, rebalance, dedup
429 and reblock commands on the specified
434 is specified this command will clean-up all
436 file systems in use, including PFS's.
437 To do this it will scan all
441 mounts, extract PFS id's, and clean-up each PFS found.
443 This command will access a snapshots
444 directory and a configuration file for each
446 creating them if necessary.
447 .Bl -tag -width indent
448 .It Nm HAMMER No version 2-
449 The configuration file is
451 in the snapshots directory which defaults to
452 .Pa <pfs>/snapshots .
453 .It Nm HAMMER No version 3+
454 The configuration file is saved in file system meta-data, see
457 The snapshots directory defaults to
458 .Pa /var/hammer/<pfs>
459 .Pa ( /var/hammer/root
463 The format of the configuration file is:
464 .Bd -literal -offset indent
465 snapshots <period> <retention-time> [any]
466 prune <period> <max-runtime>
467 rebalance <period> <max-runtime>
468 dedup <period> <max-runtime>
469 reblock <period> <max-runtime>
470 recopy <period> <max-runtime>
474 .Bd -literal -offset indent
475 snapshots 1d 60d # 0d 0d for PFS /tmp, /var/tmp, /usr/obj
483 Time is given with a suffix of
489 meaning day, hour, minute and second.
493 directive has a period of 0 and a retention time of 0
494 then snapshot generation is disabled, removal of old snapshots are
495 disabled, and prunes will use
496 .Cm prune-everything .
500 directive has a period of 0 but a non-zero retention time
501 then this command will not create any new snapshots but will remove old
502 snapshots it finds based on the retention time. This form should be
503 used on PFS masters where you are generating your own snapshot softlinks
504 manually and on PFS slaves when all you wish to do is prune away existing
505 snapshots inherited via the mirroring stream.
507 By default only snapshots in the form
508 .Ql snap- Ns Ar yyyymmdd Ns Op - Ns Ar HHMM
512 directive is specified as a third argument on the
514 config line then any softlink of the form
515 .Ql *- Ns Ar yyyymmdd Ns Op - Ns Ar HHMM
517 .Ql *. Ns Ar yyyymmdd Ns Op - Ns Ar HHMM
520 A prune max-runtime of 0 means unlimited.
522 If period hasn't passed since the previous
525 For example a day has passed when midnight is passed (localtime).
533 The default configuration file will create a daily snapshot, do a daily
534 pruning, rebalancing, deduping and reblocking run and a monthly recopy run.
535 Reblocking is defragmentation with a level of 95%,
536 and recopy is full defragmentation.
538 By default prune and rebalance operations are time limited to 5 minutes,
539 dedup and reblock operations to a bit over 5 minutes,
540 and recopy operations to a bit over 10 minutes.
541 Reblocking and recopy runs are each broken down into four separate functions:
542 btree, inodes, dirs and data.
543 Each function is time limited to the time given in the configuration file,
544 but the btree, inodes and dirs functions usually does not take very long time,
545 full defragmentation is always used for these three functions.
546 Also note that this directive will by default disable snapshots on
553 The defaults may be adjusted by modifying the configuration file.
554 The pruning and reblocking commands automatically maintain a cyclefile
555 for incremental operation.
556 If you interrupt (^C) the program the cyclefile will be updated,
558 may continue to run in the background for a few seconds until the
560 ioctl detects the interrupt.
563 PFS option can be set to use another location for the snapshots directory.
565 Work on this command is still in progress.
567 An ability to remove snapshots dynamically as the
568 file system becomes full.
570 .It Cm config Op Ar filesystem Op Ar configfile
573 Show or change configuration for
575 If zero or one arguments are specified this function dumps the current
576 configuration file to stdout.
577 Zero arguments specifies the PFS containing the current directory.
578 This configuration file is stored in file system meta-data.
579 If two arguments are specified this function installs a new config file.
583 versions less than 3 the configuration file is by default stored in
584 .Pa <pfs>/snapshots/config ,
585 but in all later versions the configuration file is stored in file system
587 .\" ==== viconfig ====
588 .It Cm viconfig Op Ar filesystem
591 Edit the configuration file and reinstall into file system meta-data when done.
592 Zero arguments specifies the PFS containing the current directory.
593 .\" ==== volume-add ====
594 .It Cm volume-add Ar device Ar filesystem
595 This command will format
597 and add all of its space to
601 All existing data contained on
603 will be destroyed by this operation!
608 file system, formatting will be denied.
609 You can overcome this sanity check
612 to erase the beginning sectors of the device.
613 Also remember that you have to specify
615 together with any other device that make up the file system,
620 .\" ==== volume-del ====
621 .It Cm volume-del Ar device Ar filesystem
622 This command will remove volume
627 Remember that you have to remove
629 from the colon-separated list in
633 .\" ==== snapshot ====
634 .It Cm snapshot Oo Ar filesystem Oc Ar snapshot-dir
635 .It Cm snapshot Ar filesystem Ar snapshot-dir Op Ar note
636 Takes a snapshot of the file system either explicitly given by
638 or implicitly derived from the
640 argument and creates a symlink in the directory provided by
642 pointing to the snapshot.
645 is not a directory, it is assumed to be a format string passed to
647 with the current time as parameter.
650 refers to an existing directory, a default format string of
652 is assumed and used as name for the newly created symlink.
654 Snapshot is a per PFS operation, so a
656 file system and each PFS in it have to be snapshot separately.
658 Example, assuming that
666 are file systems on their own, the following invocations:
667 .Bd -literal -offset indent
668 hammer snapshot /mysnapshots
670 hammer snapshot /mysnapshots/%Y-%m-%d
672 hammer snapshot /obj /mysnapshots/obj-%Y-%m-%d
674 hammer snapshot /usr /my/snaps/usr "note"
677 Would create symlinks similar to:
678 .Bd -literal -offset indent
679 /mysnapshots/snap-20080627-1210 -> /@@0x10d2cd05b7270d16
681 /mysnapshots/2008-06-27 -> /@@0x10d2cd05b7270d16
683 /mysnapshots/obj-2008-06-27 -> /obj@@0x10d2cd05b7270d16
685 /my/snaps/usr/snap-20080627-1210 -> /usr@@0x10d2cd05b7270d16
690 version 3+ file system the snapshot is also recorded in file system meta-data
691 along with the optional
697 .It Cm snap Ar path Op Ar note
700 Create a snapshot for the PFS containing
702 and create a snapshot softlink.
703 If the path specified is a
704 directory a standard snapshot softlink will be created in the directory.
705 The snapshot softlink points to the base of the mounted PFS.
706 .It Cm snaplo Ar path Op Ar note
709 Create a snapshot for the PFS containing
711 and create a snapshot softlink.
712 If the path specified is a
713 directory a standard snapshot softlink will be created in the directory.
714 The snapshot softlink points into the directory it is contained in.
715 .It Cm snapq Ar dir Op Ar note
718 Create a snapshot for the PFS containing the specified directory but do
719 not create a softlink.
720 Instead output a path which can be used to access
721 the directory via the snapshot.
723 An absolute or relative path may be specified.
724 The path will be used as-is as a prefix in the path output to stdout.
726 snap and snapshot directives the snapshot transaction id will be registered
727 in the file system meta-data.
728 .It Cm snaprm Ar path Ar ...
729 .It Cm snaprm Ar transid Ar ...
730 .It Cm snaprm Ar filesystem Ar transid Ar ...
733 Remove a snapshot given its softlink or transaction id.
734 If specifying a transaction id
735 the snapshot is removed from file system meta-data but you are responsible
736 for removing any related softlinks.
738 If a softlink path is specified the filesystem and transaction id
739 is derived from the contents of the softlink.
740 If just a transaction id is specified it is assumed to be a snapshot
741 in the HAMMER filesystem you are currently chdir'd into.
742 You can also specify the filesystem and transaction id explicitly.
743 .It Cm snapls Op Ar path ...
746 Dump the snapshot meta-data for PFSs containing each
748 listing all available snapshots and their notes.
749 If no arguments are specified snapshots for the PFS containing the
750 current directory are listed.
751 This is the definitive list of snapshots for the file system.
753 .It Cm prune Ar softlink-dir
754 Prune the file system based on previously created snapshot softlinks.
755 Pruning is the act of deleting file system history.
758 command will delete file system history such that
759 the file system state is retained for the given snapshots,
760 and all history after the latest snapshot.
761 By setting the per PFS parameter
763 history is guaranteed to be saved at least this time interval.
764 All other history is deleted.
766 The target directory is expected to contain softlinks pointing to
767 snapshots of the file systems you wish to retain.
768 The directory is scanned non-recursively and the mount points and
769 transaction ids stored in the softlinks are extracted and sorted.
770 The file system is then explicitly pruned according to what is found.
771 Cleaning out portions of the file system is as simple as removing a
772 snapshot softlink and then running the
776 As a safety measure pruning only occurs if one or more softlinks are found
779 snapshot id extension.
780 Currently the scanned softlink directory must contain softlinks pointing
784 The softlinks may specify absolute or relative paths.
785 Softlinks must use 20-character
787 transaction ids, as might be returned from
788 .Nm Cm synctid Ar filesystem .
790 Pruning is a per PFS operation, so a
792 file system and each PFS in it have to be pruned separately.
794 Note that pruning a file system may not immediately free-up space,
795 though typically some space will be freed if a large number of records are
797 The file system must be reblocked to completely recover all available space.
799 Example, lets say your that you didn't set
801 and snapshot directory contains the following links:
802 .Bd -literal -offset indent
803 lrwxr-xr-x 1 root wheel 29 May 31 17:57 snap1 ->
804 /usr/obj/@@0x10d2cd05b7270d16
806 lrwxr-xr-x 1 root wheel 29 May 31 17:58 snap2 ->
807 /usr/obj/@@0x10d2cd13f3fde98f
809 lrwxr-xr-x 1 root wheel 29 May 31 17:59 snap3 ->
810 /usr/obj/@@0x10d2cd222adee364
813 If you were to run the
815 command on this directory, then the
818 mount will be pruned to retain the above three snapshots.
819 In addition, history for modifications made to the file system older than
820 the oldest snapshot will be destroyed and history for potentially fine-grained
821 modifications made to the file system more recently than the most recent
822 snapshot will be retained.
824 If you then delete the
826 softlink and rerun the
829 history for modifications pertaining to that snapshot would be destroyed.
833 file system versions 3+ this command also scans the snapshots stored
834 in the file system meta-data and includes them in the prune.
835 .\" ==== prune-everything ====
836 .It Cm prune-everything Ar filesystem
837 This command will remove all historical records from the file system.
838 This directive is not normally used on a production system.
840 This command does not remove snapshot softlinks but will delete all
841 snapshots recorded in file system meta-data (for file system version 3+).
842 The user is responsible for deleting any softlinks.
844 Pruning is a per PFS operation, so a
846 file system and each PFS in it have to be pruned separately.
847 .\" ==== rebalance ====
848 .It Cm rebalance Ar filesystem Op Ar saturation_percentage
849 This command will rebalance the B-Tree, nodes with small number of
850 elements will be combined and element counts will be smoothed out
853 The saturation percentage is between 50% and 100%.
854 The default is 75% (the
856 suffix is not needed).
858 Rebalancing is a per PFS operation, so a
860 file system and each PFS in it have to be rebalanced separately.
862 .It Cm dedup Ar filesystem
865 Perform offline (post-process) deduplication. Deduplication occurs at
866 the block level, currently only data blocks of the same size can be
867 deduped, metadata blocks can not. The hash function used for comparing
868 data blocks is CRC-32 (CRCs are computed anyways as part of
870 data integrity features, so there's no additional overhead). Since CRC
871 is a weak hash function a byte-by-byte comparison is done before actual
872 deduping. In case of a CRC collision (two data blocks have the same CRC
873 but different contents) the checksum is upgraded to SHA-256.
877 reblocker may partially blow up (re-expand) dedup (reblocker's normal
878 operation is to reallocate every record, so it's possible for deduped
879 blocks to be re-expanded back).
881 Deduplication is a per PFS operation, so a
883 file system and each PFS in it have to be deduped separately. This also
884 means that if you have duplicated data in two different PFSs that data
885 won't be deduped, however the addition of such feature is planned.
886 .\" ==== dedup-simulate ====
887 .It Cm dedup-simulate Ar filesystem
890 Shows potential space savings (simulated dedup ratio) one can get after
893 command. If the estimated dedup ratio is greater than 1.00 you will see
894 dedup space savings. Remember that this is an estimated number, in
895 practice real dedup ratio will be slightly smaller because of
897 bigblock underflows, B-Tree locking issues and other factors.
898 .\" ==== reblock* ====
899 .It Cm reblock Ar filesystem Op Ar fill_percentage
900 .It Cm reblock-btree Ar filesystem Op Ar fill_percentage
901 .It Cm reblock-inodes Ar filesystem Op Ar fill_percentage
902 .It Cm reblock-dirs Ar filesystem Op Ar fill_percentage
903 .It Cm reblock-data Ar filesystem Op Ar fill_percentage
904 Attempt to defragment and free space for reuse by reblocking a live
907 Big-blocks cannot be reused by
909 until they are completely free.
910 This command also has the effect of reordering all elements, effectively
911 defragmenting the file system.
913 The default fill percentage is 100% and will cause the file system to be
914 completely defragmented.
915 All specified element types will be reallocated and rewritten.
916 If you wish to quickly free up space instead try specifying
917 a smaller fill percentage, such as 90% or 80% (the
919 suffix is not needed).
921 Since this command may rewrite the entire contents of the disk it is
922 best to do it incrementally from a
928 options to limit the run time.
929 The file system would thus be defragmented over long period of time.
931 It is recommended that separate invocations be used for each data type.
932 B-Tree nodes, inodes, and directories are typically the most important
933 elements needing defragmentation.
934 Data can be defragmented over a longer period of time.
936 Reblocking is a per PFS operation, so a
938 file system and each PFS in it have to be reblocked separately.
939 .\" ==== pfs-status ====
940 .It Cm pfs-status Ar dirpath ...
941 Retrieve the mirroring configuration parameters for the specified
943 file systems or pseudo-filesystems (PFS's).
944 .\" ==== pfs-master ====
945 .It Cm pfs-master Ar dirpath Op Ar options
946 Create a pseudo-filesystem (PFS) inside a
949 Up to 65535 such file systems can be created.
950 Each PFS uses an independent inode numbering space making it suitable
951 for use as a replication source or target.
955 directive creates a PFS that you can read, write, and use as a mirroring
958 It is recommended to use a
960 mount to access a PFS, for more information see
962 .\" ==== pfs-slave ====
963 .It Cm pfs-slave Ar dirpath Op Ar options
964 Create a pseudo-filesystem (PFS) inside a
967 Up to 65535 such file systems can be created.
968 Each PFS uses an independent inode numbering space making it suitable
969 for use as a replication source or target.
973 directive creates a PFS that you can use as a mirroring target.
974 You will not be able to access a slave PFS until you have completed the
975 first mirroring operation with it as the target (its root directory will
976 not exist until then).
978 Access to the pfs-slave via the special softlink, as described in the
983 dynamically modify the snapshot transaction id by returning a dynamic result
988 A PFS can only be truly destroyed with the
991 Removing the softlink will not destroy the underlying PFS.
993 It is recommended to use a
995 mount to access a PFS, for more information see
997 .\" ==== pfs-update ====
998 .It Cm pfs-update Ar dirpath Op Ar options
999 Update the configuration parameters for an existing
1001 file system or pseudo-filesystem.
1002 Options that may be specified:
1003 .Bl -tag -width indent
1004 .It Cm sync-beg-tid= Ns Ar 0x16llx
1005 This is the automatic snapshot access starting transaction id for
1007 This parameter is normally updated automatically by the
1011 It is important to note that accessing a mirroring slave
1012 with a transaction id greater than the last fully synchronized transaction
1013 id can result in an unreliable snapshot since you will be accessing
1014 data that is still undergoing synchronization.
1016 Manually modifying this field is dangerous and can result in a broken mirror.
1017 .It Cm sync-end-tid= Ns Ar 0x16llx
1018 This is the current synchronization point for mirroring slaves.
1019 This parameter is normally updated automatically by the
1023 Manually modifying this field is dangerous and can result in a broken mirror.
1024 .It Cm shared-uuid= Ns Ar uuid
1025 Set the shared UUID for this file system.
1026 All mirrors must have the same shared UUID.
1027 For safety purposes the
1029 directives will refuse to operate on a target with a different shared UUID.
1031 Changing the shared UUID on an existing, non-empty mirroring target,
1032 including an empty but not completely pruned target,
1033 can lead to corruption of the mirroring target.
1034 .It Cm unique-uuid= Ns Ar uuid
1035 Set the unique UUID for this file system.
1036 This UUID should not be used anywhere else,
1037 even on exact copies of the file system.
1038 .It Cm label= Ns Ar string
1039 Set a descriptive label for this file system.
1040 .It Cm snapshots= Ns Ar string
1041 Specify the snapshots directory which
1044 will use to manage this PFS.
1045 .Bl -tag -width indent
1046 .It Nm HAMMER No version 2-
1047 The snapshots directory does not need to be configured for
1048 PFS masters and will default to
1049 .Pa <pfs>/snapshots .
1051 PFS slaves are mirroring slaves so you cannot configure a snapshots
1052 directory on the slave itself to be managed by the slave's machine.
1053 In fact, the slave will likely have a
1055 sub-directory mirrored
1056 from the master, but that directory contains the configuration the master
1057 is using for its copy of the file system, not the configuration that we
1058 want to use for our slave.
1060 It is recommended that
1061 .Pa <fs>/var/slaves/<name>
1062 be configured for a PFS slave, where
1068 is an appropriate label.
1069 .It Nm HAMMER No version 3+
1070 The snapshots directory does not need to be configured for PFS masters or
1072 The snapshots directory defaults to
1073 .Pa /var/hammer/<pfs>
1074 .Pa ( /var/hammer/root
1078 You can control snapshot retention on your slave independent of the master.
1079 .It Cm snapshots-clear
1082 directory path for this PFS.
1083 .It Cm prune-min= Ns Ar N Ns Cm d
1084 .It Cm prune-min= Ns Oo Ar N Ns Cm d/ Oc Ns \
1085 Ar hh Ns Op Cm \&: Ns Ar mm Ns Op Cm \&: Ns Ar ss
1086 Set the minimum fine-grained data retention period.
1088 always retains fine-grained history up to the most recent snapshot.
1089 You can extend the retention period further by specifying a non-zero
1091 Any snapshot softlinks within the retention period are ignored
1092 for the purposes of pruning (the fine grained history is retained).
1093 Number of days, hours, minutes and seconds are given as
1098 Because the transaction id in the snapshot softlink cannot be used
1099 to calculate a timestamp,
1101 uses the earlier of the
1105 field of the softlink to
1106 determine which snapshots fall within the retention period.
1107 Users must be sure to retain one of these two fields when manipulating
1110 .\" ==== pfs-upgrade ====
1111 .It Cm pfs-upgrade Ar dirpath
1112 Upgrade a PFS from slave to master operation.
1113 The PFS will be rolled back to the current end synchronization transaction id
1114 (removing any partial synchronizations), and will then become writable.
1118 currently supports only single masters and using
1119 this command can easily result in file system corruption
1120 if you don't know what you are doing.
1122 This directive will refuse to run if any programs have open descriptors
1123 in the PFS, including programs chdir'd into the PFS.
1124 .\" ==== pfs-downgrade ====
1125 .It Cm pfs-downgrade Ar dirpath
1126 Downgrade a master PFS from master to slave operation.
1127 The PFS becomes read-only and access will be locked to its
1130 This directive will refuse to run if any programs have open descriptors
1131 in the PFS, including programs chdir'd into the PFS.
1132 .\" ==== pfs-destroy ====
1133 .It Cm pfs-destroy Ar dirpath
1134 This permanently destroys a PFS.
1136 This directive will refuse to run if any programs have open descriptors
1137 in the PFS, including programs chdir'd into the PFS.
1138 .\" ==== mirror-read ====
1139 .It Cm mirror-read Ar filesystem Op Ar begin-tid
1140 Generate a mirroring stream to stdout.
1141 The stream ends when the transaction id space has been exhausted.
1142 .\" ==== mirror-read-stream ====
1143 .It Cm mirror-read-stream Ar filesystem Op Ar begin-tid
1144 Generate a mirroring stream to stdout.
1145 Upon completion the stream is paused until new data is synced to the
1148 Operation continues until the pipe is broken.
1151 command for more details.
1152 .\" ==== mirror-write ====
1153 .It Cm mirror-write Ar filesystem
1154 Take a mirroring stream on stdin.
1156 This command will fail if the
1158 configuration field for the two file systems do not match.
1161 command for more details.
1163 If the target PFS does not exist this command will ask you whether
1164 you want to create a compatible PFS slave for the target or not.
1165 .\" ==== mirror-dump ====
1171 to dump an ASCII representation of the mirroring stream.
1172 .\" ==== mirror-copy ====
1173 .\".It Cm mirror-copy Ar [[user@]host:]filesystem [[user@]host:]filesystem
1174 .It Cm mirror-copy \
1175 Oo Oo Ar user Ns Cm @ Oc Ns Ar host Ns Cm \&: Oc Ns Ar filesystem \
1176 Oo Oo Ar user Ns Cm @ Oc Ns Ar host Ns Cm \&: Oc Ns Ar filesystem
1177 This is a shortcut which pipes a
1182 If a remote host specification is made the program forks a
1188 on the appropriate host.
1189 The source may be a master or slave PFS, and the target must be a slave PFS.
1191 This command also establishes full duplex communication and turns on
1192 the 2-way protocol feature
1194 which automatically negotiates transaction id
1195 ranges without having to use a cyclefile.
1196 If the operation completes successfully the target PFS's
1199 Note that you must re-chdir into the target PFS to see the updated information.
1200 If you do not you will still be in the previous snapshot.
1202 If the target PFS does not exist this command will ask you whether
1203 you want to create a compatible PFS slave for the target or not.
1204 .\" ==== mirror-stream ====
1205 .\".It Cm mirror-stream Ar [[user@]host:]filesystem [[user@]host:]filesystem
1206 .It Cm mirror-stream \
1207 Oo Oo Ar user Ns Cm @ Oc Ns Ar host Ns Cm \&: Oc Ns Ar filesystem \
1208 Oo Oo Ar user Ns Cm @ Oc Ns Ar host Ns Cm \&: Oc Ns Ar filesystem
1209 This is a shortcut which pipes a
1210 .Cm mirror-read-stream
1214 This command works similarly to
1216 but does not exit after the initial mirroring completes.
1217 The mirroring operation will resume as changes continue to be made to the
1219 The command is commonly used with
1223 options to keep the mirroring target in sync with the source on a continuing
1226 If the pipe is broken the command will automatically retry after sleeping
1228 The time slept will be 15 seconds plus the time given in the
1232 This command also detects the initial-mirroring case and spends some
1233 time scanning the B-Tree to find good break points, allowing the initial
1234 bulk mirroring operation to be broken down into 100MB pieces.
1235 This means that the user can kill and restart the operation and it will
1236 not have to start from scratch once it has gotten past the first chunk.
1239 option may be used to change the size of pieces and the
1241 option may be used to disable this feature and perform an initial bulk
1243 .\" ==== version ====
1244 .It Cm version Ar filesystem
1245 This command returns the
1247 file system version for the specified
1249 as well as the range of versions supported in the kernel.
1252 option may be used to remove the summary at the end.
1253 .\" ==== version-upgrade ====
1254 .It Cm version-upgrade Ar filesystem Ar version Op Cm force
1255 This command upgrades the
1260 Once upgraded a file system may not be downgraded.
1261 If you wish to upgrade a file system to a version greater or equal to the
1262 work-in-progress version number you must specify the
1265 Use of WIP versions should be relegated to testing and may require wiping
1266 the file system as development progresses, even though the WIP version might
1270 This command operates on the entire
1272 file system and is not a per PFS operation.
1273 All PFS's will be affected.
1274 .Bl -tag -width indent
1277 default version, first
1282 New directory entry layout.
1283 This version is using a new directory hash key.
1286 New snapshot management, using file system meta-data for saving
1287 configuration file and snapshots (transaction ids etc.).
1288 Also default snapshots directory has changed.
1292 New undo/redo/flush, giving HAMMER a much faster sync and fsync.
1295 .Sh PSEUDO-FILESYSTEM (PFS) NOTES
1296 The root of a PFS is not hooked into the primary
1298 file system as a directory.
1301 creates a special softlink called
1303 (exactly 10 characters long) in the primary
1307 then modifies the contents of the softlink as read by
1309 and thus what you see with an
1311 command or if you were to
1314 If the PFS is a master the link reflects the current state of the PFS.
1315 If the PFS is a slave the link reflects the last completed snapshot, and the
1316 contents of the link will change when the next snapshot is completed, and
1321 utility employs numerous safeties to reduce user foot-shooting.
1324 directive requires that the target be configured as a slave and that the
1326 field of the mirroring source and target match.
1327 .Sh UPGRADE INSTRUCTIONS HAMMER V1 TO V2
1328 This upgrade changes the way directory entries are stored.
1329 It is possible to upgrade a V1 file system to V2 in place, but
1330 directories created prior to the upgrade will continue to use
1333 Note that the slave mirroring code in the target kernel had bugs in
1334 V1 which can create an incompatible root directory on the slave.
1337 master created after the upgrade with a
1339 slave created prior to the upgrade.
1341 Any directories created after upgrading will use a new layout.
1342 .Sh UPGRADE INSTRUCTIONS HAMMER V2 TO V3
1343 This upgrade adds meta-data elements to the B-Tree.
1344 It is possible to upgrade a V2 file system to V3 in place.
1345 After issuing the upgrade be sure to run a
1348 to perform post-upgrade tasks.
1350 After making this upgrade running a
1355 directory for each PFS mount into
1356 .Pa /var/hammer/<pfs> .
1359 root mount will migrate
1362 .Pa /var/hammer/root .
1363 Migration occurs only once and only if you have not specified
1364 a snapshots directory in the PFS configuration.
1365 If you have specified a snapshots directory in the PFS configuration no
1366 automatic migration will occur.
1368 For slaves, if you desire, you can migrate your snapshots
1369 config to the new location manually and then clear the
1370 snapshot directory configuration in the slave PFS.
1371 The new snapshots hierarchy is designed to work with
1372 both master and slave PFSs equally well.
1374 In addition, the old config file will be moved to file system meta-data,
1375 editable via the new
1379 The old config file will be deleted.
1380 Migration occurs only once.
1382 The V3 file system has new
1384 directives for creating snapshots.
1385 All snapshot directives, including the original, will create
1386 meta-data entries for the snapshots and the pruning code will
1387 automatically incorporate these entries into its list and
1388 expire them the same way it expires softlinks.
1389 If you by accident blow away your snapshot softlinks you can use the
1391 directive to get a definitive list from the file system meta-data and
1392 regenerate them from that list.
1397 to backup file systems your scripts may be using the
1399 directive to generate transaction ids.
1400 This directive does not create a snapshot.
1401 You will have to modify your scripts to use the
1403 directive to generate the linkbuf for the softlink you create, or
1404 use one of the other
1409 directive will continue to work as expected and in V3 it will also
1410 record the snapshot transaction id in file system meta-data.
1411 You may also want to make use of the new
1413 tag for the meta-data.
1416 If you used to remove snapshot softlinks with
1418 you should probably start using the
1420 directive instead to also remove the related meta-data.
1421 The pruning code scans the meta-data so just removing the
1422 softlink is not sufficient.
1423 .Sh UPGRADE INSTRUCTIONS HAMMER V3 TO V4
1424 This upgrade changes undo/flush, giving faster sync.
1425 It is possible to upgrade a V3 file system to V4 in place.
1426 This upgrade reformats the UNDO FIFO (typically 1GB), so upgrade might take
1427 a minute or two depending.
1429 Version 4 allows the UNDO FIFO to be flushed without also having
1430 to flush the volume header, removing 2 of the 4 disk syncs typically
1433 and removing 1 of the 2 disk syncs typically
1434 required for a flush sequence.
1435 Version 4 also implements the REDO log (see below) which is capable
1436 of fsync()ing with either one disk flush or zero disk flushes.
1437 .Sh FSYNC FLUSH MODES
1439 implements five different fsync flush modes via the
1440 .Va vfs.hammer.fsync_mode
1443 version 4+ file systems.
1447 fsync mode 3 is set by default.
1448 REDO operation and recovery is enabled by default.
1449 .Bl -tag -width indent
1451 Full synchronous fsync semantics without REDO.
1454 will not generate REDOs.
1457 will completely sync
1458 the data and meta-data and double-flush the FIFO, including
1459 issuing two disk synchronization commands.
1460 The data is guaranteed
1461 to be on the media as of when
1464 Needless to say, this is slow.
1466 Relaxed asynchronous fsync semantics without REDO.
1468 This mode works the same as mode 0 except the last disk synchronization
1469 command is not issued.
1470 It is faster than mode 0 but not even remotely
1471 close to the speed you get with mode 2 or mode 3.
1473 Note that there is no chance of meta-data corruption when using this
1474 mode, it simply means that the data you wrote and then
1476 might not have made it to the media if the storage system crashes at a bad
1480 Full synchronous fsync semantics using REDO.
1481 NOTE: If not running
1482 a HAMMER version 4 filesystem or later mode 0 is silently used.
1485 will generate REDOs in the UNDO/REDO FIFO based on a heuristic.
1486 If this is sufficient to satisfy the
1488 operation the blocks
1489 will be written out and
1491 will wait for the I/Os to complete,
1492 and then followup with a disk sync command to guarantee the data
1493 is on the media before returning.
1494 This is slower than mode 3 and can result in significant disk or
1495 SSDs overheads, though not as bad as mode 0 or mode 1.
1498 Relaxed asynchronous fsync semantics using REDO.
1499 NOTE: If not running
1500 a HAMMER version 4 filesystem or later mode 1 is silently used.
1503 will generate REDOs in the UNDO/REDO FIFO based on a heuristic.
1504 If this is sufficient to satisfy the
1506 operation the blocks
1507 will be written out and
1509 will wait for the I/Os to complete,
1512 issue a disk synchronization command.
1514 Note that there is no chance of meta-data corruption when using this
1515 mode, it simply means that the data you wrote and then
1518 not have made it to the media if the storage system crashes at a bad
1521 This mode is the fastest production fsyncing mode available.
1522 This mode is equivalent to how the UFS fsync in the
1532 This mode is primarily designed
1533 for testing and should not be used on a production system.
1535 .Sh RESTORING FROM A SNAPSHOT BACKUP
1536 You restore a snapshot by copying it over to live, but there is a caveat.
1537 The mtime and atime fields for files accessed via a snapshot is locked
1538 to the ctime in order to keep the snapshot consistent, because neither
1539 mtime nor atime changes roll any history.
1541 In order to avoid unnecessary copying it is recommended that you use
1545 when doing the copyback. Also make sure you traverse the snapshot softlink
1546 by appending a ".", as in "<snapshotpath>/.", and you match up the directory
1551 If the following environment variables exist, they will be used by:
1552 .Bl -tag -width ".Ev EDITOR"
1554 The editor program specified in the variable
1556 will be invoked instead of the default editor, which is
1564 .Bl -tag -width ".It Pa <fs>/var/slaves/<name>" -compact
1565 .It Pa <pfs>/snapshots
1566 default per PFS snapshots directory
1569 .It Pa /var/hammer/<pfs>
1570 default per PFS snapshots directory (not root)
1573 .It Pa /var/hammer/root
1574 default snapshots directory for root directory
1577 .It Pa <snapshots>/config
1584 .It Pa <fs>/var/slaves/<name>
1585 recommended slave PFS snapshots directory
1594 .Xr periodic.conf 5 ,
1595 .Xr mount_hammer 8 ,
1601 utility first appeared in
1604 .An Matthew Dillon Aq dillon@backplane.com