1.\" Copyright (c) 2007 The DragonFly Project. All rights reserved. 2.\" 3.\" This code is derived from software contributed to The DragonFly Project 4.\" by Matthew Dillon <dillon@backplane.com> 5.\" 6.\" Redistribution and use in source and binary forms, with or without 7.\" modification, are permitted provided that the following conditions 8.\" are met: 9.\" 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 15.\" distribution. 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. 19.\" 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 31.\" SUCH DAMAGE. 32.\" 33.Dd April 19, 2011 34.Dt HAMMER 8 35.Os 36.Sh NAME 37.Nm hammer 38.Nd HAMMER file system utility 39.Sh SYNOPSIS 40.Nm 41.Fl h 42.Nm 43.Op Fl 2BFqrvXy 44.Op Fl b Ar bandwidth 45.Op Fl C Ar cachesize Ns Op Ns Cm \&: Ns Ar readahead 46.Op Fl c Ar cyclefile 47.Op Fl f Ar blkdevs 48.\" .Op Fl s Ar linkpath 49.Op Fl i Ar delay 50.Op Fl p Ar ssh-port 51.Op Fl S Ar splitsize 52.Op Fl t Ar seconds 53.Ar command 54.Op Ar argument ... 55.Sh DESCRIPTION 56This manual page documents the 57.Nm 58utility which provides miscellaneous functions related to managing a 59.Nm HAMMER 60file system. 61For a general introduction to the 62.Nm HAMMER 63file system, its features, and 64examples on how to set up and maintain one, see 65.Xr HAMMER 5 . 66.Pp 67The options are as follows: 68.Bl -tag -width indent 69.It Fl 2 70Tell the mirror commands to use a 2-way protocol, which allows 71automatic negotiation of transaction id ranges. 72This option is automatically enabled by the 73.Cm mirror-copy 74command. 75.It Fl B 76Bulk transfer. 77.Cm Mirror-stream 78will not attempt to break-up large initial bulk transfers into smaller 79pieces. 80This can save time but if the link is lost in the middle of the 81initial bulk transfer you will have to start over from scratch. 82For more information see the 83.Fl S 84option. 85.It Fl b Ar bandwidth 86Specify a bandwidth limit in bytes per second for mirroring streams. 87This option is typically used to prevent batch mirroring operations from 88loading down the machine. 89The bandwidth may be suffixed with 90.Cm k , m , 91or 92.Cm g 93to specify values in kilobytes, megabytes, and gigabytes per second. 94If no suffix is specified, bytes per second is assumed. 95.Pp 96Unfortunately this is only applicable to the pre-compression bandwidth 97when compression is used, so a better solution would probably be to 98use a 99.Xr ipfw 8 100pipe or a 101.Xr pf 4 102queue. 103.It Fl C Ar cachesize Ns Op Ns Cm \&: Ns Ar readahead 104Set the memory cache size for any raw 105.Tn I/O . 106The default is 16MB. 107A suffix of 108.Cm k 109for kilobytes and 110.Cm m 111for megabytes is allowed, 112else the cache size is specified in bytes. 113.Pp 114The read-behind/read-ahead defaults to 4 115.Nm HAMMER 116blocks. 117.Pp 118This option is typically only used with diagnostic commands 119as kernel-supported commands will use the kernel's buffer cache. 120.It Fl c Ar cyclefile 121When pruning, rebalancing or reblocking you can tell the utility 122to start at the object id stored in the specified file. 123If the file does not exist 124.Nm 125will start at the beginning. 126If 127.Nm 128is told to run for a specific period of time 129.Pq Fl t 130and is unable to complete the operation it will write out 131the current object id so the next run can pick up where it left off. 132If 133.Nm 134runs to completion it will delete 135.Ar cyclefile . 136.It Fl F 137Force operation. 138E.g.\& 139.Cm cleanup 140will not check that time period has elapsed if this option is given. 141.It Fl f Ar blkdevs 142Specify the volumes making up a 143.Nm HAMMER 144file system. 145.Ar Blkdevs 146is a colon-separated list of devices, each specifying a 147.Nm HAMMER 148volume. 149.It Fl h 150Show usage. 151.It Fl i Ar delay 152Specify delay in seconds for 153.Cm mirror-read-stream . 154When maintaining a streaming mirroring this option specifies the 155minimum delay after a batch ends before the next batch is allowed 156to start. 157The default is five seconds. 158.It Fl p Ar ssh-port 159Pass the 160.Fl p Ar ssh-port 161option to 162.Xr ssh 1 163when using a remote 164specification for the source and/or destination. 165.It Fl q 166Decrease verboseness. 167May be specified multiple times. 168.It Fl r 169Specify recursion for those commands which support it. 170.It Fl S Ar splitsize 171Specify the bulk splitup size in bytes for mirroring streams. 172When a 173.Cm mirror-stream 174is first started 175.Nm 176will do an initial run-through of the data to calculate good 177transaction ids to cut up the bulk transfers, creating 178restart points in case the stream is interrupted. 179If we don't do this and the stream is interrupted it might 180have to start all over again. 181The default is a 182.Ar splitsize 183of 4GB. 184.Pp 185At the moment the run-through is disk-bandwidth-heavy but some 186future version will limit the run-through to just the B-Tree 187records and not the record data. 188.Pp 189The splitsize may be suffixed with 190.Cm k , m , 191or 192.Cm g 193to specify values in kilobytes, megabytes, or gigabytes. 194If no suffix is specified, bytes is assumed. 195.Pp 196When mirroring very large filesystems the minimum recommended 197split size is 4GB. 198A small split size may wind up generating a great deal of overhead 199but very little actual incremental data and is not recommended. 200.It Fl t Ar seconds 201Specify timeout in seconds. 202When pruning, rebalancing, reblocking or mirror-reading 203you can tell the utility to stop after a certain period of time. 204A value of 0 means unlimited. 205This option is used along with the 206.Fl c Ar cyclefile 207option to prune, rebalance or reblock incrementally. 208.It Fl v 209Increase verboseness. 210May be specified multiple times. 211.It Fl X 212Enable compression for any remote ssh specifications. 213This option is typically used with the mirroring directives. 214.It Fl y 215Force 216.Dq yes 217for interactive questions. 218.El 219.Pp 220The commands are as follows: 221.Bl -tag -width indent 222.\" ==== synctid ==== 223.It Cm synctid Ar filesystem Op Cm quick 224Generate a guaranteed, formal 64-bit transaction id representing the 225current state of the specified 226.Nm HAMMER 227file system. 228The file system will be synced to the media. 229.Pp 230If the 231.Cm quick 232keyword is specified the file system will be soft-synced, meaning that a 233crash might still undo the state of the file system as of the transaction 234id returned but any new modifications will occur after the returned 235transaction id as expected. 236.Pp 237This operation does not create a snapshot. 238It is meant to be used 239to track temporary fine-grained changes to a subset of files and 240will only remain valid for 241.Ql @@ 242access purposes for the 243.Cm prune-min 244period configured for the PFS. 245If you desire a real snapshot then the 246.Cm snapq 247directive may be what you are looking for. 248.\" ==== bstats ==== 249.It Cm bstats Op Ar interval 250Output 251.Nm HAMMER 252B-Tree statistics until interrupted. 253Pause 254.Ar interval 255seconds between each display. 256The default interval is one second. 257.\" ==== iostats ==== 258.It Cm iostats Op Ar interval 259Output 260.Nm HAMMER 261.Tn I/O 262statistics until interrupted. 263Pause 264.Ar interval 265seconds between each display. 266The default interval is one second. 267.\" ==== history ==== 268.It Cm history Ns Oo Cm @ Ns Ar offset Ns Oo Cm \&, Ns Ar length Oc Oc Ar path ... 269Show the modification history for inode and data of 270.Nm HAMMER 271files. 272If 273.Ar offset 274is given history is shown for data block at given offset, 275otherwise history is shown for inode. 276If 277.Fl v 278is specified 279.Ar length 280data bytes at given offset are dumped for each version, 281default is 32. 282.Pp 283For each 284.Ar path 285this directive shows object id and sync status, 286and for each object version it shows transaction id and time stamp. 287Files has to exist for this directive to be applicable, 288to track inodes which has been deleted or renamed see 289.Xr undo 1 . 290.\" ==== blockmap ==== 291.It Cm blockmap 292Dump the blockmap for the file system. 293The 294.Nm HAMMER 295blockmap is two-layer 296blockmap representing the maximum possible file system size of 1 Exabyte. 297Needless to say the second layer is only present for blocks which exist. 298.Nm HAMMER Ns 's 299blockmap represents 8-Megabyte blocks, called big-blocks. 300Each big-block has an append 301point, a free byte count, and a typed zone id which allows content to be 302reverse engineered to some degree. 303.Pp 304In 305.Nm HAMMER 306allocations are essentially appended to a selected big-block using 307the append offset and deducted from the free byte count. 308When space is freed the free byte count is adjusted but 309.Nm HAMMER 310does not track holes in big-blocks for reallocation. 311A big-block must be completely freed, either 312through normal file system operations or through reblocking, before 313it can be reused. 314.Pp 315Data blocks can be shared by deducting the space used from the free byte 316count for each shared references. 317This means the free byte count can legally go negative. 318.Pp 319This command needs the 320.Fl f Ar blkdevs 321option. 322.\" ==== checkmap ==== 323.It Cm checkmap 324Check the blockmap allocation count. 325.Nm 326will scan the B-Tree, collect allocation information, and 327construct a blockmap in-memory. 328It will then check that blockmap against the on-disk blockmap. 329.Pp 330This command needs the 331.Fl f Ar blkdevs 332option. 333.\" ==== show ==== 334.It Cm show Op Ar localization Ns Op Cm \&: Ns Ar object_id 335Dump the B-Tree. 336By default this command will validate all B-Tree 337linkages and CRCs, including data CRCs, and will report the most verbose 338information it can dig up. 339Any errors will show up with a 340.Ql B 341in column 1 along with various 342other error flags. 343.Pp 344If you specify 345.Ar localization 346or 347.Ar localization Ns Cm \&: Ns Ar object_id 348the dump will 349search for the key printing nodes as it recurses down, and then 350will iterate forwards. 351These fields are specified in HEX. 352Note that the pfsid is the top 16 bits of the 32-bit localization 353field so PFS #1 would be 00010000. 354.Pp 355If you use 356.Fl q 357the command will report less information about the inode contents. 358.Pp 359If you use 360.Fl qq 361the command will not report the content of the inode or other typed 362data at all. 363.Pp 364If you use 365.Fl qqq 366the command will not report volume header information, big-block fill 367ratios, mirror transaction ids, or report or check data CRCs. 368B-Tree CRCs and linkages are still checked. 369.Pp 370This command needs the 371.Fl f Ar blkdevs 372option. 373.\" ==== show-undo ==== 374.It Cm show-undo 375.Nm ( HAMMER 376VERSION 4+) 377Dump the UNDO/REDO map. 378.Pp 379This command needs the 380.Fl f Ar blkdevs 381option. 382.\" .It Ar blockmap 383.\" Dump the B-Tree, record, large-data, and small-data blockmaps, showing 384.\" physical block assignments and free space percentages. 385.\" ==== recover ==== 386.It Cm recover Ar targetdir 387Recover data from a corrupted 388.Nm HAMMER 389filesystem. 390This is a low level command which operates on the filesystem image and 391attempts to locate and recover files from a corrupted filesystem. 392The entire image is scanned linearly looking for B-Tree nodes. 393Any node 394found which passes its CRC test is scanned for file, inode, and directory 395fragments and the target directory is populated with the resulting data. 396files and directories in the target directory are initially named after 397the object id and are renamed as fragmentary information is processed. 398.Pp 399This command keeps track of filename/object_id translations and may eat a 400considerably amount of memory while operating. 401.Pp 402This command is literally the last line of defense when it comes to 403recovering data from a dead filesystem. 404.Pp 405This command needs the 406.Fl f Ar blkdevs 407option. 408.\" ==== namekey1 ==== 409.It Cm namekey1 Ar filename 410Generate a 411.Nm HAMMER 41264-bit directory hash for the specified file name, using 413the original directory hash algorithm in version 1 of the file system. 414The low 32 bits are used as an iterator for hash collisions and will be 415output as 0. 416.\" ==== namekey2 ==== 417.It Cm namekey2 Ar filename 418Generate a 419.Nm HAMMER 42064-bit directory hash for the specified file name, using 421the new directory hash algorithm in version 2 of the file system. 422The low 32 bits are still used as an iterator but will start out containing 423part of the hash key. 424.\" ==== namekey32 ==== 425.It Cm namekey32 Ar filename 426Generate the top 32 bits of a 427.Nm HAMMER 42864 bit directory hash for the specified file name. 429.\" ==== info ==== 430.It Cm info 431Show extended information about 432.Nm HAMMER 433file systems. 434The information is divided into sections: 435.Bl -tag -width indent 436.It Volume identification 437General information, like the label of the 438.Nm HAMMER 439filesystem, the number of volumes it contains, the FSID, and the 440.Nm HAMMER 441version being used. 442.It Big block information 443Big block statistics, such as total, used, reserved and free big blocks. 444.It Space information 445Information about space used on the filesystem. 446Currently total size, used, reserved and free space are displayed. 447.It PFS information 448Basic information about the PFSs currently present on a 449.Nm HAMMER 450filesystem. 451.Pp 452.Dq PFS ID 453is the ID of the PFS, with 0 being the root PFS. 454.Dq Snaps 455is the current snapshot count on the PFS. 456.Dq Mounted on 457displays the mount point of the PFS is currently mounted on (if any). 458.El 459.\" ==== cleanup ==== 460.It Cm cleanup Op Ar filesystem ... 461This is a meta-command which executes snapshot, prune, rebalance, dedup 462and reblock commands on the specified 463.Nm HAMMER 464file systems. 465If no 466.Ar filesystem 467is specified this command will clean-up all 468.Nm HAMMER 469file systems in use, including PFS's. 470To do this it will scan all 471.Nm HAMMER 472and 473.Nm null 474mounts, extract PFS id's, and clean-up each PFS found. 475.Pp 476This command will access a snapshots 477directory and a configuration file for each 478.Ar filesystem , 479creating them if necessary. 480.Bl -tag -width indent 481.It Nm HAMMER No version 2- 482The configuration file is 483.Pa config 484in the snapshots directory which defaults to 485.Pa <pfs>/snapshots . 486.It Nm HAMMER No version 3+ 487The configuration file is saved in file system meta-data, see 488.Nm 489.Cm config . 490The snapshots directory defaults to 491.Pa /var/hammer/<pfs> 492.Pa ( /var/hammer/root 493for root mount). 494.El 495.Pp 496The format of the configuration file is: 497.Bd -literal -offset indent 498snapshots <period> <retention-time> [any] 499prune <period> <max-runtime> 500rebalance <period> <max-runtime> 501dedup <period> <max-runtime> 502reblock <period> <max-runtime> 503recopy <period> <max-runtime> 504.Ed 505.Pp 506Defaults are: 507.Bd -literal -offset indent 508snapshots 1d 60d # 0d 0d for PFS /tmp, /var/tmp, /usr/obj 509prune 1d 5m 510rebalance 1d 5m 511dedup 1d 5m 512reblock 1d 5m 513recopy 30d 10m 514.Ed 515.Pp 516Time is given with a suffix of 517.Cm d , 518.Cm h , 519.Cm m 520or 521.Cm s 522meaning day, hour, minute and second. 523.Pp 524If the 525.Cm snapshots 526directive has a period of 0 and a retention time of 0 527then snapshot generation is disabled, removal of old snapshots are 528disabled, and prunes will use 529.Cm prune-everything . 530.Pp 531If the 532.Cm snapshots 533directive has a period of 0 but a non-zero retention time 534then this command will not create any new snapshots but will remove old 535snapshots it finds based on the retention time. 536This form should be 537used on PFS masters where you are generating your own snapshot softlinks 538manually and on PFS slaves when all you wish to do is prune away existing 539snapshots inherited via the mirroring stream. 540.Pp 541By default only snapshots in the form 542.Ql snap- Ns Ar yyyymmdd Ns Op - Ns Ar HHMM 543are processed. 544If the 545.Cm any 546directive is specified as a third argument on the 547.Cm snapshots 548config line then any softlink of the form 549.Ql *- Ns Ar yyyymmdd Ns Op - Ns Ar HHMM 550or 551.Ql *. Ns Ar yyyymmdd Ns Op - Ns Ar HHMM 552will be processed. 553.Pp 554A period of 0 for prune, rebalance, dedup, reblock or recopy disables the directive. 555A max-runtime of 0 means unlimited. 556.Pp 557If period hasn't passed since the previous 558.Cm cleanup 559run nothing is done. 560For example a day has passed when midnight is passed (localtime). 561If the 562.Fl F 563flag is given the period is ignored. 564By default, 565.Dx 566is set up to run 567.Nm Cm cleanup 568nightly via 569.Xr periodic 8 . 570.Pp 571The default configuration file will create a daily snapshot, do a daily 572pruning, rebalancing, deduping and reblocking run and a monthly recopy run. 573Reblocking is defragmentation with a level of 95%, 574and recopy is full defragmentation. 575.Pp 576By default prune, dedup and rebalance operations are time limited to 5 minutes, 577and reblock operations to a bit over 5 minutes, 578and recopy operations to a bit over 10 minutes. 579Reblocking and recopy runs are each broken down into four separate functions: 580btree, inodes, dirs and data. 581Each function is time limited to the time given in the configuration file, 582but the btree, inodes and dirs functions usually does not take very long time, 583full defragmentation is always used for these three functions. 584Also note that this directive will by default disable snapshots on 585the following PFS's: 586.Pa /tmp , 587.Pa /var/tmp 588and 589.Pa /usr/obj . 590.Pp 591The defaults may be adjusted by modifying the configuration file. 592The pruning and reblocking commands automatically maintain a cyclefile 593for incremental operation. 594If you interrupt (^C) the program the cyclefile will be updated, 595but a sub-command 596may continue to run in the background for a few seconds until the 597.Nm HAMMER 598ioctl detects the interrupt. 599The 600.Cm snapshots 601PFS option can be set to use another location for the snapshots directory. 602.Pp 603Work on this command is still in progress. 604Expected additions: 605An ability to remove snapshots dynamically as the 606file system becomes full. 607.\" ==== config ==== 608.It Cm config Op Ar filesystem Op Ar configfile 609.Nm ( HAMMER 610VERSION 3+) 611Show or change configuration for 612.Ar filesystem . 613If zero or one arguments are specified this function dumps the current 614configuration file to stdout. 615Zero arguments specifies the PFS containing the current directory. 616This configuration file is stored in file system meta-data. 617If two arguments are specified this function installs a new config file. 618.Pp 619In 620.Nm HAMMER 621versions less than 3 the configuration file is by default stored in 622.Pa <pfs>/snapshots/config , 623but in all later versions the configuration file is stored in file system 624meta-data. 625.\" ==== viconfig ==== 626.It Cm viconfig Op Ar filesystem 627.Nm ( HAMMER 628VERSION 3+) 629Edit the configuration file and reinstall into file system meta-data when done. 630Zero arguments specifies the PFS containing the current directory. 631.\" ==== volume-add ==== 632.It Cm volume-add Ar device Ar filesystem 633Add volume 634.Ar device 635to 636.Ar filesystem . 637This will format 638.Ar device 639and add all of its space to 640.Ar filesystem . 641A 642.Nm HAMMER 643file system can use up to 256 volumes. 644.Pp 645.Em NOTE! 646All existing data contained on 647.Ar device 648will be destroyed by this operation! 649If 650.Ar device 651contains a valid 652.Nm HAMMER 653file system, formatting will be denied. 654You can overcome this sanity check by using 655.Xr dd 1 656to erase the beginning sectors of the device. 657.Pp 658Remember that you have to specify 659.Ar device , 660together with any other device that make up the file system, 661colon-separated to 662.Pa /etc/fstab 663and 664.Xr mount_hammer 8 . 665If 666.Ar filesystem 667is root file system, also remember to add 668.Ar device 669to 670.Va vfs.root.mountfrom 671in 672.Pa /boot/loader.conf , 673see 674.Xr loader 8 . 675.\" ==== volume-del ==== 676.It Cm volume-del Ar device Ar filesystem 677Remove volume 678.Ar device 679from 680.Ar filesystem . 681.Pp 682Remember that you have to remove 683.Ar device 684from the colon-separated list in 685.Pa /etc/fstab 686and 687.Xr mount_hammer 8 . 688If 689.Ar filesystem 690is root file system, also remember to remove 691.Ar device 692from 693.Va vfs.root.mountfrom 694in 695.Pa /boot/loader.conf , 696see 697.Xr loader 8 . 698.\" ==== volume-list ==== 699.It Cm volume-list Ar filesystem 700List the volumes that make up 701.Ar filesystem . 702.\" ==== snapshot ==== 703.It Cm snapshot Oo Ar filesystem Oc Ar snapshot-dir 704.It Cm snapshot Ar filesystem Ar snapshot-dir Op Ar note 705Take a snapshot of the file system either explicitly given by 706.Ar filesystem 707or implicitly derived from the 708.Ar snapshot-dir 709argument and creates a symlink in the directory provided by 710.Ar snapshot-dir 711pointing to the snapshot. 712If 713.Ar snapshot-dir 714is not a directory, it is assumed to be a format string passed to 715.Xr strftime 3 716with the current time as parameter. 717If 718.Ar snapshot-dir 719refers to an existing directory, a default format string of 720.Ql snap-%Y%m%d-%H%M 721is assumed and used as name for the newly created symlink. 722.Pp 723Snapshot is a per PFS operation, so each PFS in a 724.Nm HAMMER 725file system have to be snapshot separately. 726.Pp 727Example, assuming that 728.Pa /mysnapshots 729is on file system 730.Pa / 731and that 732.Pa /obj 733and 734.Pa /usr 735are file systems on their own, the following invocations: 736.Bd -literal -offset indent 737hammer snapshot /mysnapshots 738 739hammer snapshot /mysnapshots/%Y-%m-%d 740 741hammer snapshot /obj /mysnapshots/obj-%Y-%m-%d 742 743hammer snapshot /usr /my/snaps/usr "note" 744.Ed 745.Pp 746Would create symlinks similar to: 747.Bd -literal -offset indent 748/mysnapshots/snap-20080627-1210 -> /@@0x10d2cd05b7270d16 749 750/mysnapshots/2008-06-27 -> /@@0x10d2cd05b7270d16 751 752/mysnapshots/obj-2008-06-27 -> /obj@@0x10d2cd05b7270d16 753 754/my/snaps/usr/snap-20080627-1210 -> /usr@@0x10d2cd05b7270d16 755.Ed 756.Pp 757When run on a 758.Nm HAMMER 759version 3+ file system the snapshot is also recorded in file system meta-data 760along with the optional 761.Ar note . 762See the 763.Cm snapls 764directive. 765.\" ==== snap* ==== 766.It Cm snap Ar path Op Ar note 767.Nm ( HAMMER 768VERSION 3+) 769Create a snapshot for the PFS containing 770.Ar path 771and create a snapshot softlink. 772If the path specified is a 773directory a standard snapshot softlink will be created in the directory. 774The snapshot softlink points to the base of the mounted PFS. 775.It Cm snaplo Ar path Op Ar note 776.Nm ( HAMMER 777VERSION 3+) 778Create a snapshot for the PFS containing 779.Ar path 780and create a snapshot softlink. 781If the path specified is a 782directory a standard snapshot softlink will be created in the directory. 783The snapshot softlink points into the directory it is contained in. 784.It Cm snapq Ar dir Op Ar note 785.Nm ( HAMMER 786VERSION 3+) 787Create a snapshot for the PFS containing the specified directory but do 788not create a softlink. 789Instead output a path which can be used to access 790the directory via the snapshot. 791.Pp 792An absolute or relative path may be specified. 793The path will be used as-is as a prefix in the path output to stdout. 794As with the other 795snap and snapshot directives the snapshot transaction id will be registered 796in the file system meta-data. 797.It Cm snaprm Ar path Ar ... 798.It Cm snaprm Ar transaction_id Ar ... 799.It Cm snaprm Ar filesystem Ar transaction_id Ar ... 800.Nm ( HAMMER 801VERSION 3+) 802Remove a snapshot given its softlink or transaction id. 803If specifying a transaction id 804the snapshot is removed from file system meta-data but you are responsible 805for removing any related softlinks. 806.Pp 807If a softlink path is specified the filesystem and transaction id 808is derived from the contents of the softlink. 809If just a transaction id is specified it is assumed to be a snapshot in the 810.Nm HAMMER 811filesystem you are currently chdir'd into. 812You can also specify the filesystem and transaction id explicitly. 813.It Cm snapls Op Ar path ... 814.Nm ( HAMMER 815VERSION 3+) 816Dump the snapshot meta-data for PFSs containing each 817.Ar path 818listing all available snapshots and their notes. 819If no arguments are specified snapshots for the PFS containing the 820current directory are listed. 821This is the definitive list of snapshots for the file system. 822.\" ==== prune ==== 823.It Cm prune Ar softlink-dir 824Prune the file system based on previously created snapshot softlinks. 825Pruning is the act of deleting file system history. 826The 827.Cm prune 828command will delete file system history such that 829the file system state is retained for the given snapshots, 830and all history after the latest snapshot. 831By setting the per PFS parameter 832.Cm prune-min , 833history is guaranteed to be saved at least this time interval. 834All other history is deleted. 835.Pp 836The target directory is expected to contain softlinks pointing to 837snapshots of the file systems you wish to retain. 838The directory is scanned non-recursively and the mount points and 839transaction ids stored in the softlinks are extracted and sorted. 840The file system is then explicitly pruned according to what is found. 841Cleaning out portions of the file system is as simple as removing a 842snapshot softlink and then running the 843.Cm prune 844command. 845.Pp 846As a safety measure pruning only occurs if one or more softlinks are found 847containing the 848.Ql @@ 849snapshot id extension. 850Currently the scanned softlink directory must contain softlinks pointing 851to a single 852.Nm HAMMER 853mount. 854The softlinks may specify absolute or relative paths. 855Softlinks must use 20-character 856.Ql @@0x%016llx 857transaction ids, as might be returned from 858.Nm Cm synctid Ar filesystem . 859.Pp 860Pruning is a per PFS operation, so each PFS in a 861.Nm HAMMER 862file system have to be pruned separately. 863.Pp 864Note that pruning a file system may not immediately free-up space, 865though typically some space will be freed if a large number of records are 866pruned out. 867The file system must be reblocked to completely recover all available space. 868.Pp 869Example, lets say your that you didn't set 870.Cm prune-min , 871and snapshot directory contains the following links: 872.Bd -literal -offset indent 873lrwxr-xr-x 1 root wheel 29 May 31 17:57 snap1 -> 874/usr/obj/@@0x10d2cd05b7270d16 875 876lrwxr-xr-x 1 root wheel 29 May 31 17:58 snap2 -> 877/usr/obj/@@0x10d2cd13f3fde98f 878 879lrwxr-xr-x 1 root wheel 29 May 31 17:59 snap3 -> 880/usr/obj/@@0x10d2cd222adee364 881.Ed 882.Pp 883If you were to run the 884.Cm prune 885command on this directory, then the 886.Nm HAMMER 887.Pa /usr/obj 888mount will be pruned to retain the above three snapshots. 889In addition, history for modifications made to the file system older than 890the oldest snapshot will be destroyed and history for potentially fine-grained 891modifications made to the file system more recently than the most recent 892snapshot will be retained. 893.Pp 894If you then delete the 895.Pa snap2 896softlink and rerun the 897.Cm prune 898command, 899history for modifications pertaining to that snapshot would be destroyed. 900.Pp 901In 902.Nm HAMMER 903file system versions 3+ this command also scans the snapshots stored 904in the file system meta-data and includes them in the prune. 905.\" ==== prune-everything ==== 906.It Cm prune-everything Ar filesystem 907Remove all historical records from 908.Ar filesystem . 909Use this directive with caution on PFSs where you intend to use history. 910.Pp 911This command does not remove snapshot softlinks but will delete all 912snapshots recorded in file system meta-data (for file system version 3+). 913The user is responsible for deleting any softlinks. 914.Pp 915Pruning is a per PFS operation, so each PFS in a 916.Nm HAMMER 917file system have to be pruned separately. 918.\" ==== rebalance ==== 919.It Cm rebalance Ar filesystem Op Ar saturation_percentage 920Rebalance the B-Tree, nodes with small number of 921elements will be combined and element counts will be smoothed out 922between nodes. 923.Pp 924The saturation percentage is between 50% and 100%. 925The default is 85% (the 926.Sq % 927suffix is not needed). 928.Pp 929Rebalancing is a per PFS operation, so each PFS in a 930.Nm HAMMER 931file system have to be rebalanced separately. 932.\" ==== dedup ==== 933.It Cm dedup Ar filesystem 934.Nm ( HAMMER 935VERSION 5+) 936Perform offline (post-process) deduplication. 937Deduplication occurs at 938the block level, currently only data blocks of the same size can be 939deduped, metadata blocks can not. 940The hash function used for comparing 941data blocks is CRC-32 (CRCs are computed anyways as part of 942.Nm HAMMER 943data integrity features, so there's no additional overhead). 944Since CRC is a weak hash function a byte-by-byte comparison is done 945before actual deduping. 946In case of a CRC collision (two data blocks have the same CRC 947but different contents) the checksum is upgraded to SHA-256. 948.Pp 949Currently 950.Nm HAMMER 951reblocker may partially blow up (re-expand) dedup (reblocker's normal 952operation is to reallocate every record, so it's possible for deduped 953blocks to be re-expanded back). 954.Pp 955Deduplication is a per PFS operation, so each PFS in a 956.Nm HAMMER 957file system have to be deduped separately. 958This also 959means that if you have duplicated data in two different PFSs that data 960won't be deduped, however the addition of such feature is planned. 961.\" ==== dedup-simulate ==== 962.It Cm dedup-simulate Ar filesystem 963Shows potential space savings (simulated dedup ratio) one can get after 964running 965.Cm dedup 966command. 967If the estimated dedup ratio is greater than 1.00 you will see 968dedup space savings. 969Remember that this is an estimated number, in 970practice real dedup ratio will be slightly smaller because of 971.Nm HAMMER 972bigblock underflows, B-Tree locking issues and other factors. 973.Pp 974Note that deduplication currently works only on bulk data so if you 975try to run 976.Cm dedup-simulate 977or 978.Cm dedup 979commands on a PFS that contains metadata only (directory entries, 980softlinks) you will get a 0.00 dedup ratio. 981.\" ==== reblock* ==== 982.It Cm reblock Ar filesystem Op Ar fill_percentage 983.It Cm reblock-btree Ar filesystem Op Ar fill_percentage 984.It Cm reblock-inodes Ar filesystem Op Ar fill_percentage 985.It Cm reblock-dirs Ar filesystem Op Ar fill_percentage 986.It Cm reblock-data Ar filesystem Op Ar fill_percentage 987Attempt to defragment and free space for reuse by reblocking a live 988.Nm HAMMER 989file system. 990Big-blocks cannot be reused by 991.Nm HAMMER 992until they are completely free. 993This command also has the effect of reordering all elements, effectively 994defragmenting the file system. 995.Pp 996The default fill percentage is 100% and will cause the file system to be 997completely defragmented. 998All specified element types will be reallocated and rewritten. 999If you wish to quickly free up space instead try specifying 1000a smaller fill percentage, such as 90% or 80% (the 1001.Sq % 1002suffix is not needed). 1003.Pp 1004Since this command may rewrite the entire contents of the disk it is 1005best to do it incrementally from a 1006.Xr cron 8 1007job along with the 1008.Fl c Ar cyclefile 1009and 1010.Fl t Ar seconds 1011options to limit the run time. 1012The file system would thus be defragmented over long period of time. 1013.Pp 1014It is recommended that separate invocations be used for each data type. 1015B-Tree nodes, inodes, and directories are typically the most important 1016elements needing defragmentation. 1017Data can be defragmented over a longer period of time. 1018.Pp 1019Reblocking is a per PFS operation, so each PFS in a 1020.Nm HAMMER 1021file system have to be reblocked separately. 1022.\" ==== pfs-status ==== 1023.It Cm pfs-status Ar dirpath ... 1024Retrieve the mirroring configuration parameters for the specified 1025.Nm HAMMER 1026file systems or pseudo-filesystems (PFS's). 1027.\" ==== pfs-master ==== 1028.It Cm pfs-master Ar dirpath Op Ar options 1029Create a pseudo-filesystem (PFS) inside a 1030.Nm HAMMER 1031file system. 1032Up to 65536 PFSs can be created. 1033Each PFS uses an independent inode numbering space making it suitable 1034for replication. 1035.Pp 1036The 1037.Cm pfs-master 1038directive creates a PFS that you can read, write, and use as a mirroring 1039source. 1040.Pp 1041A PFS can only be truly destroyed with the 1042.Cm pfs-destroy 1043directive. 1044Removing the softlink will not destroy the underlying PFS. 1045.Pp 1046A PFS can only be created in the root PFS (PFS# 0), 1047not in a PFS created by 1048.Cm pfs-master 1049or 1050.Cm pfs-slave 1051(PFS# >0). 1052.Pp 1053It is recommended that 1054.Ar dirpath 1055is of the form 1056.Pa <fs>/pfs/<name> 1057(i.e.\& located in 1058.Pa pfs 1059directory at root of 1060.Nm HAMMER 1061file system). 1062.Pp 1063It is recommended to use a 1064.Nm null 1065mount to access a PFS, except for root PFS, for more information see 1066.Xr HAMMER 5 . 1067.\" ==== pfs-slave ==== 1068.It Cm pfs-slave Ar dirpath Op Ar options 1069Create a pseudo-filesystem (PFS) inside a 1070.Nm HAMMER 1071file system. 1072Up to 65536 PFSs can be created. 1073Each PFS uses an independent inode numbering space making it suitable 1074for replication. 1075.Pp 1076The 1077.Cm pfs-slave 1078directive creates a PFS that you can use as a mirroring source or target. 1079You will not be able to access a slave PFS until you have completed the 1080first mirroring operation with it as the target (its root directory will 1081not exist until then). 1082.Pp 1083Access to the pfs-slave via the special softlink, as described in the 1084.Sx PFS NOTES 1085below, allows 1086.Nm HAMMER 1087to 1088dynamically modify the snapshot transaction id by returning a dynamic result 1089from 1090.Xr readlink 2 1091calls. 1092.Pp 1093A PFS can only be truly destroyed with the 1094.Cm pfs-destroy 1095directive. 1096Removing the softlink will not destroy the underlying PFS. 1097.Pp 1098A PFS can only be created in the root PFS (PFS# 0), 1099not in a PFS created by 1100.Cm pfs-master 1101or 1102.Cm pfs-slave 1103(PFS# >0). 1104.Pp 1105It is recommended that 1106.Ar dirpath 1107is of the form 1108.Pa <fs>/pfs/<name> 1109(i.e.\& located in 1110.Pa pfs 1111directory at root of 1112.Nm HAMMER 1113file system). 1114.Pp 1115It is recommended to use a 1116.Nm null 1117mount to access a PFS, except for root PFS, for more information see 1118.Xr HAMMER 5 . 1119.\" ==== pfs-update ==== 1120.It Cm pfs-update Ar dirpath Op Ar options 1121Update the configuration parameters for an existing 1122.Nm HAMMER 1123file system or pseudo-filesystem. 1124Options that may be specified: 1125.Bl -tag -width indent 1126.It Cm sync-beg-tid= Ns Ar 0x16llx 1127This is the automatic snapshot access starting transaction id for 1128mirroring slaves. 1129This parameter is normally updated automatically by the 1130.Cm mirror-write 1131directive. 1132.Pp 1133It is important to note that accessing a mirroring slave 1134with a transaction id greater than the last fully synchronized transaction 1135id can result in an unreliable snapshot since you will be accessing 1136data that is still undergoing synchronization. 1137.Pp 1138Manually modifying this field is dangerous and can result in a broken mirror. 1139.It Cm sync-end-tid= Ns Ar 0x16llx 1140This is the current synchronization point for mirroring slaves. 1141This parameter is normally updated automatically by the 1142.Cm mirror-write 1143directive. 1144.Pp 1145Manually modifying this field is dangerous and can result in a broken mirror. 1146.It Cm shared-uuid= Ns Ar uuid 1147Set the shared UUID for this file system. 1148All mirrors must have the same shared UUID. 1149For safety purposes the 1150.Cm mirror-write 1151directives will refuse to operate on a target with a different shared UUID. 1152.Pp 1153Changing the shared UUID on an existing, non-empty mirroring target, 1154including an empty but not completely pruned target, 1155can lead to corruption of the mirroring target. 1156.It Cm unique-uuid= Ns Ar uuid 1157Set the unique UUID for this file system. 1158This UUID should not be used anywhere else, 1159even on exact copies of the file system. 1160.It Cm label= Ns Ar string 1161Set a descriptive label for this file system. 1162.It Cm snapshots= Ns Ar string 1163Specify the snapshots directory which 1164.Nm 1165.Cm cleanup 1166will use to manage this PFS. 1167.Bl -tag -width indent 1168.It Nm HAMMER No version 2- 1169The snapshots directory does not need to be configured for 1170PFS masters and will default to 1171.Pa <pfs>/snapshots . 1172.Pp 1173PFS slaves are mirroring slaves so you cannot configure a snapshots 1174directory on the slave itself to be managed by the slave's machine. 1175In fact, the slave will likely have a 1176.Pa snapshots 1177sub-directory mirrored 1178from the master, but that directory contains the configuration the master 1179is using for its copy of the file system, not the configuration that we 1180want to use for our slave. 1181.Pp 1182It is recommended that 1183.Pa <fs>/var/slaves/<name> 1184be configured for a PFS slave, where 1185.Pa <fs> 1186is the base 1187.Nm HAMMER 1188file system, and 1189.Pa <name> 1190is an appropriate label. 1191.It Nm HAMMER No version 3+ 1192The snapshots directory does not need to be configured for PFS masters or 1193slaves. 1194The snapshots directory defaults to 1195.Pa /var/hammer/<pfs> 1196.Pa ( /var/hammer/root 1197for root mount). 1198.El 1199.Pp 1200You can control snapshot retention on your slave independent of the master. 1201.It Cm snapshots-clear 1202Zero out the 1203.Cm snapshots 1204directory path for this PFS. 1205.It Cm prune-min= Ns Ar N Ns Cm d 1206.It Cm prune-min= Ns Oo Ar N Ns Cm d/ Oc Ns \ 1207Ar hh Ns Op Cm \&: Ns Ar mm Ns Op Cm \&: Ns Ar ss 1208Set the minimum fine-grained data retention period. 1209.Nm HAMMER 1210always retains fine-grained history up to the most recent snapshot. 1211You can extend the retention period further by specifying a non-zero 1212pruning minimum. 1213Any snapshot softlinks within the retention period are ignored 1214for the purposes of pruning (i.e.\& the fine grained history is retained). 1215Number of days, hours, minutes and seconds are given as 1216.Ar N , hh , mm 1217and 1218.Ar ss . 1219.Pp 1220Because the transaction id in the snapshot softlink cannot be used 1221to calculate a timestamp, 1222.Nm HAMMER 1223uses the earlier of the 1224.Fa st_ctime 1225or 1226.Fa st_mtime 1227field of the softlink to 1228determine which snapshots fall within the retention period. 1229Users must be sure to retain one of these two fields when manipulating 1230the softlink. 1231.El 1232.\" ==== pfs-upgrade ==== 1233.It Cm pfs-upgrade Ar dirpath 1234Upgrade a PFS from slave to master operation. 1235The PFS will be rolled back to the current end synchronization transaction id 1236(removing any partial synchronizations), and will then become writable. 1237.Pp 1238.Em WARNING! 1239.Nm HAMMER 1240currently supports only single masters and using 1241this command can easily result in file system corruption 1242if you don't know what you are doing. 1243.Pp 1244This directive will refuse to run if any programs have open descriptors 1245in the PFS, including programs chdir'd into the PFS. 1246.\" ==== pfs-downgrade ==== 1247.It Cm pfs-downgrade Ar dirpath 1248Downgrade a master PFS from master to slave operation. 1249The PFS becomes read-only and access will be locked to its 1250.Cm sync-end-tid . 1251.Pp 1252This directive will refuse to run if any programs have open descriptors 1253in the PFS, including programs chdir'd into the PFS. 1254.\" ==== pfs-destroy ==== 1255.It Cm pfs-destroy Ar dirpath 1256This permanently destroys a PFS. 1257.Pp 1258This directive will refuse to run if any programs have open descriptors 1259in the PFS, including programs chdir'd into the PFS. 1260As safety measure the 1261.Fl y 1262flag have no effect on this directive. 1263.\" ==== mirror-read ==== 1264.It Cm mirror-read Ar filesystem Op Ar begin-tid 1265Generate a mirroring stream to stdout. 1266The stream ends when the transaction id space has been exhausted. 1267.Ar filesystem 1268may be a master or slave PFS. 1269.\" ==== mirror-read-stream ==== 1270.It Cm mirror-read-stream Ar filesystem Op Ar begin-tid 1271Generate a mirroring stream to stdout. 1272Upon completion the stream is paused until new data is synced to the 1273.Ar filesystem , 1274then resumed. 1275Operation continues until the pipe is broken. 1276See the 1277.Cm mirror-stream 1278command for more details. 1279.\" ==== mirror-write ==== 1280.It Cm mirror-write Ar filesystem 1281Take a mirroring stream on stdin. 1282.Ar filesystem 1283must be a slave PFS. 1284.Pp 1285This command will fail if the 1286.Cm shared-uuid 1287configuration field for the two file systems do not match. 1288See the 1289.Cm mirror-copy 1290command for more details. 1291.Pp 1292If the target PFS does not exist this command will ask you whether 1293you want to create a compatible PFS slave for the target or not. 1294.\" ==== mirror-dump ==== 1295.It Cm mirror-dump 1296A 1297.Cm mirror-read 1298can be piped into a 1299.Cm mirror-dump 1300to dump an ASCII representation of the mirroring stream. 1301.\" ==== mirror-copy ==== 1302.\".It Cm mirror-copy Ar [[user@]host:]filesystem [[user@]host:]filesystem 1303.It Cm mirror-copy \ 1304Oo Oo Ar user Ns Cm @ Oc Ns Ar host Ns Cm \&: Oc Ns Ar filesystem \ 1305Oo Oo Ar user Ns Cm @ Oc Ns Ar host Ns Cm \&: Oc Ns Ar filesystem 1306This is a shortcut which pipes a 1307.Cm mirror-read 1308command to a 1309.Cm mirror-write 1310command. 1311If a remote host specification is made the program forks a 1312.Xr ssh 1 1313and execs the 1314.Cm mirror-read 1315and/or 1316.Cm mirror-write 1317on the appropriate host. 1318The source may be a master or slave PFS, and the target must be a slave PFS. 1319.Pp 1320This command also establishes full duplex communication and turns on 1321the 2-way protocol feature 1322.Fl ( 2 ) 1323which automatically negotiates transaction id 1324ranges without having to use a cyclefile. 1325If the operation completes successfully the target PFS's 1326.Cm sync-end-tid 1327will be updated. 1328Note that you must re-chdir into the target PFS to see the updated information. 1329If you do not you will still be in the previous snapshot. 1330.Pp 1331If the target PFS does not exist this command will ask you whether 1332you want to create a compatible PFS slave for the target or not. 1333.\" ==== mirror-stream ==== 1334.\".It Cm mirror-stream Ar [[user@]host:]filesystem [[user@]host:]filesystem 1335.It Cm mirror-stream \ 1336Oo Oo Ar user Ns Cm @ Oc Ns Ar host Ns Cm \&: Oc Ns Ar filesystem \ 1337Oo Oo Ar user Ns Cm @ Oc Ns Ar host Ns Cm \&: Oc Ns Ar filesystem 1338This is a shortcut which pipes a 1339.Cm mirror-read-stream 1340command to a 1341.Cm mirror-write 1342command. 1343This command works similarly to 1344.Cm mirror-copy 1345but does not exit after the initial mirroring completes. 1346The mirroring operation will resume as changes continue to be made to the 1347source. 1348The command is commonly used with 1349.Fl i Ar delay 1350and 1351.Fl b Ar bandwidth 1352options to keep the mirroring target in sync with the source on a continuing 1353basis. 1354.Pp 1355If the pipe is broken the command will automatically retry after sleeping 1356for a short while. 1357The time slept will be 15 seconds plus the time given in the 1358.Fl i 1359option. 1360.Pp 1361This command also detects the initial-mirroring case and spends some 1362time scanning the B-Tree to find good break points, allowing the initial 1363bulk mirroring operation to be broken down into 4GB pieces. 1364This means that the user can kill and restart the operation and it will 1365not have to start from scratch once it has gotten past the first chunk. 1366The 1367.Fl S 1368option may be used to change the size of pieces and the 1369.Fl B 1370option may be used to disable this feature and perform an initial bulk 1371transfer instead. 1372.\" ==== version ==== 1373.It Cm version Ar filesystem 1374This command returns the 1375.Nm HAMMER 1376file system version for the specified 1377.Ar filesystem 1378as well as the range of versions supported in the kernel. 1379The 1380.Fl q 1381option may be used to remove the summary at the end. 1382.\" ==== version-upgrade ==== 1383.It Cm version-upgrade Ar filesystem Ar version Op Cm force 1384Upgrade the 1385.Nm HAMMER 1386.Ar filesystem 1387to the specified 1388.Ar version . 1389Once upgraded a file system may not be downgraded. 1390If you wish to upgrade a file system to a version greater or equal to the 1391work-in-progress (WIP) version number you must specify the 1392.Cm force 1393directive. 1394Use of WIP versions should be relegated to testing and may require wiping 1395the file system as development progresses, even though the WIP version might 1396not change. 1397.Pp 1398.Em NOTE! 1399This command operates on the entire 1400.Nm HAMMER 1401file system and is not a per PFS operation. 1402All PFS's will be affected. 1403.Bl -tag -width indent 1404.It 1 1405.Dx 2.0 1406default version, first 1407.Nm HAMMER 1408release. 1409.It 2 1410.Dx 2.3 . 1411New directory entry layout. 1412This version is using a new directory hash key. 1413.It 3 1414.Dx 2.5 . 1415New snapshot management, using file system meta-data for saving 1416configuration file and snapshots (transaction ids etc.). 1417Also default snapshots directory has changed. 1418.It 4 1419.Dx 2.6 1420default version. 1421New undo/redo/flush, giving 1422.Nm HAMMER 1423a much faster sync and fsync. 1424.It 5 1425.Dx 2.9 . 1426Deduplication support. 1427.It 6 1428.Dx 2.9 . 1429Directory hash ALG1. 1430Tends to maintain inode number / directory name entry ordering better 1431for files after minor renaming. 1432.El 1433.El 1434.Sh PSEUDO-FILESYSTEM (PFS) NOTES 1435The root of a PFS is not hooked into the primary 1436.Nm HAMMER 1437file system as a directory. 1438Instead, 1439.Nm HAMMER 1440creates a special softlink called 1441.Ql @@PFS%05d 1442(exactly 10 characters long) in the primary 1443.Nm HAMMER 1444file system. 1445.Nm HAMMER 1446then modifies the contents of the softlink as read by 1447.Xr readlink 2 , 1448and thus what you see with an 1449.Nm ls 1450command or if you were to 1451.Nm cd 1452into the link. 1453If the PFS is a master the link reflects the current state of the PFS. 1454If the PFS is a slave the link reflects the last completed snapshot, and the 1455contents of the link will change when the next snapshot is completed, and 1456so forth. 1457.Pp 1458The 1459.Nm 1460utility employs numerous safeties to reduce user foot-shooting. 1461The 1462.Cm mirror-copy 1463directive requires that the target be configured as a slave and that the 1464.Cm shared-uuid 1465field of the mirroring source and target match. 1466.Sh DOUBLE_BUFFER MODE 1467There is a limit to the number of vnodes the kernel can cache, and because 1468file buffers are associated with a vnode the related data cache can get 1469blown away when operating on large numbers of files even if the system has 1470sufficient memory to hold the file data. 1471.Pp 1472If you turn on 1473.Nm HAMMER Ns 's 1474double buffer mode by setting the 1475.Xr sysctl 8 1476node 1477.Va vfs.hammer.double_buffer 1478to 1 1479.Nm HAMMER 1480will cache file data via the block device and copy it into the per-file 1481buffers as needed. The data will be double-cached at least until the 1482buffer cache throws away the file buffer. 1483This mode is typically used in conjunction with 1484.Xr swapcache 8 1485when 1486.Va vm.swapcache.data_enable 1487is turned on in order to prevent unnecessary re-caching of file data 1488due to vnode recycling. 1489The swapcache will save the cached VM pages related to 1490.Nm HAMMER Ns 's 1491block 1492device (which doesn't recycle unless you umount the filesystem) instead 1493of the cached VM pages backing the file vnodes. 1494.Pp 1495Double buffering should also be turned on if live dedup is enabled via 1496.Va vfs.hammer.live_dedup . 1497This is because the live dedup must validate the contents of a potential 1498duplicate file block and it must run through the block device to do that 1499and not the file vnode. 1500If double buffering is not enabled then live dedup will create extra disk 1501reads to validate potential data duplicates. 1502.Sh UPGRADE INSTRUCTIONS HAMMER V1 TO V2 1503This upgrade changes the way directory entries are stored. 1504It is possible to upgrade a V1 file system to V2 in place, but 1505directories created prior to the upgrade will continue to use 1506the old layout. 1507.Pp 1508Note that the slave mirroring code in the target kernel had bugs in 1509V1 which can create an incompatible root directory on the slave. 1510Do not mix a 1511.Nm HAMMER 1512master created after the upgrade with a 1513.Nm HAMMER 1514slave created prior to the upgrade. 1515.Pp 1516Any directories created after upgrading will use a new layout. 1517.Sh UPGRADE INSTRUCTIONS HAMMER V2 TO V3 1518This upgrade adds meta-data elements to the B-Tree. 1519It is possible to upgrade a V2 file system to V3 in place. 1520After issuing the upgrade be sure to run a 1521.Nm 1522.Cm cleanup 1523to perform post-upgrade tasks. 1524.Pp 1525After making this upgrade running a 1526.Nm 1527.Cm cleanup 1528will move the 1529.Pa <pfs>/snapshots 1530directory for each PFS mount into 1531.Pa /var/hammer/<pfs> . 1532A 1533.Nm HAMMER 1534root mount will migrate 1535.Pa /snapshots 1536into 1537.Pa /var/hammer/root . 1538Migration occurs only once and only if you have not specified 1539a snapshots directory in the PFS configuration. 1540If you have specified a snapshots directory in the PFS configuration no 1541automatic migration will occur. 1542.Pp 1543For slaves, if you desire, you can migrate your snapshots 1544config to the new location manually and then clear the 1545snapshot directory configuration in the slave PFS. 1546The new snapshots hierarchy is designed to work with 1547both master and slave PFSs equally well. 1548.Pp 1549In addition, the old config file will be moved to file system meta-data, 1550editable via the new 1551.Nm 1552.Cm viconfig 1553directive. 1554The old config file will be deleted. 1555Migration occurs only once. 1556.Pp 1557The V3 file system has new 1558.Cm snap* 1559directives for creating snapshots. 1560All snapshot directives, including the original, will create 1561meta-data entries for the snapshots and the pruning code will 1562automatically incorporate these entries into its list and 1563expire them the same way it expires softlinks. 1564If you by accident blow away your snapshot softlinks you can use the 1565.Cm snapls 1566directive to get a definitive list from the file system meta-data and 1567regenerate them from that list. 1568.Pp 1569.Em WARNING! 1570If you are using 1571.Nm 1572to backup file systems your scripts may be using the 1573.Cm synctid 1574directive to generate transaction ids. 1575This directive does not create a snapshot. 1576You will have to modify your scripts to use the 1577.Cm snapq 1578directive to generate the linkbuf for the softlink you create, or 1579use one of the other 1580.Cm snap* 1581directives. 1582The older 1583.Cm snapshot 1584directive will continue to work as expected and in V3 it will also 1585record the snapshot transaction id in file system meta-data. 1586You may also want to make use of the new 1587.Ar note 1588tag for the meta-data. 1589.Pp 1590.Em WARNING! 1591If you used to remove snapshot softlinks with 1592.Nm rm 1593you should probably start using the 1594.Cm snaprm 1595directive instead to also remove the related meta-data. 1596The pruning code scans the meta-data so just removing the 1597softlink is not sufficient. 1598.Sh UPGRADE INSTRUCTIONS HAMMER V3 TO V4 1599This upgrade changes undo/flush, giving faster sync. 1600It is possible to upgrade a V3 file system to V4 in place. 1601This upgrade reformats the UNDO/REDO FIFO (typically 1GB), 1602so upgrade might take a minute or two depending. 1603.Pp 1604Version 4 allows the UNDO/REDO FIFO to be flushed without also having 1605to flush the volume header, removing 2 of the 4 disk syncs typically 1606required for an 1607.Fn fsync 1608and removing 1 of the 2 disk syncs typically 1609required for a flush sequence. 1610Version 4 also implements the REDO log (see 1611.Sx FSYNC FLUSH MODES 1612below) which is capable 1613of fsync()ing with either one disk flush or zero disk flushes. 1614.Sh UPGRADE INSTRUCTIONS HAMMER V4 TO V5 1615This upgrade brings in deduplication support. 1616It is possible to upgrade a V4 file system to V5 in place. 1617Technically it makes the layer2 1618.Va bytes_free 1619field a signed value instead of unsigned, allowing it to go negative. 1620A version 5 filesystem is required for dedup operation. 1621.Sh UPGRADE INSTRUCTIONS HAMMER V5 TO V6 1622It is possible to upgrade a V5 file system to V6 in place. 1623.Sh FSYNC FLUSH MODES 1624.Nm HAMMER 1625implements five different fsync flush modes via the 1626.Va vfs.hammer.fsync_mode 1627sysctl, for 1628.Nm HAMMER 1629version 4+ file systems. 1630.Pp 1631As of 1632.Dx 2.6 1633fsync mode 3 is set by default. 1634REDO operation and recovery is enabled by default. 1635.Bl -tag -width indent 1636.It mode 0 1637Full synchronous fsync semantics without REDO. 1638.Pp 1639.Nm HAMMER 1640will not generate REDOs. 1641A 1642.Fn fsync 1643will completely sync 1644the data and meta-data and double-flush the FIFO, including 1645issuing two disk synchronization commands. 1646The data is guaranteed 1647to be on the media as of when 1648.Fn fsync 1649returns. 1650Needless to say, this is slow. 1651.It mode 1 1652Relaxed asynchronous fsync semantics without REDO. 1653.Pp 1654This mode works the same as mode 0 except the last disk synchronization 1655command is not issued. 1656It is faster than mode 0 but not even remotely 1657close to the speed you get with mode 2 or mode 3. 1658.Pp 1659Note that there is no chance of meta-data corruption when using this 1660mode, it simply means that the data you wrote and then 1661.Fn fsync Ns 'd 1662might not have made it to the media if the storage system crashes at a bad 1663time. 1664.Pp 1665.It mode 2 1666Full synchronous fsync semantics using REDO. 1667NOTE: If not running a 1668.Nm HAMMER 1669version 4 filesystem or later mode 0 is silently used. 1670.Pp 1671.Nm HAMMER 1672will generate REDOs in the UNDO/REDO FIFO based on a heuristic. 1673If this is sufficient to satisfy the 1674.Fn fsync 1675operation the blocks will be written out and 1676.Nm HAMMER 1677will wait for the I/Os to complete, 1678and then followup with a disk sync command to guarantee the data 1679is on the media before returning. 1680This is slower than mode 3 and can result in significant disk or 1681SSDs overheads, though not as bad as mode 0 or mode 1. 1682.Pp 1683.It mode 3 1684Relaxed asynchronous fsync semantics using REDO. 1685NOTE: If not running a 1686.Nm HAMMER 1687version 4 filesystem or later mode 1 is silently used. 1688.Pp 1689.Nm HAMMER 1690will generate REDOs in the UNDO/REDO FIFO based on a heuristic. 1691If this is sufficient to satisfy the 1692.Fn fsync 1693operation the blocks 1694will be written out and 1695.Nm HAMMER 1696will wait for the I/Os to complete, 1697but will 1698.Em NOT 1699issue a disk synchronization command. 1700.Pp 1701Note that there is no chance of meta-data corruption when using this 1702mode, it simply means that the data you wrote and then 1703.Fn fsync Ns 'd 1704might 1705not have made it to the media if the storage system crashes at a bad 1706time. 1707.Pp 1708This mode is the fastest production fsyncing mode available. 1709This mode is equivalent to how the UFS fsync in the 1710.Bx Ns s 1711operates. 1712.Pp 1713.It mode 4 1714fsync is ignored. 1715.Pp 1716Calls to 1717.Fn fsync 1718will be ignored. 1719This mode is primarily designed 1720for testing and should not be used on a production system. 1721.El 1722.Sh RESTORING FROM A SNAPSHOT BACKUP 1723You restore a snapshot by copying it over to live, but there is a caveat. 1724The mtime and atime fields for files accessed via a snapshot is locked 1725to the ctime in order to keep the snapshot consistent, because neither 1726mtime nor atime changes roll any history. 1727.Pp 1728In order to avoid unnecessary copying it is recommended that you use 1729.Nm cpdup 1730.Fl VV 1731.Fl v 1732when doing the copyback. 1733Also make sure you traverse the snapshot softlink by appending a ".", 1734as in "<snapshotpath>/.", and you match up the directory properly. 1735.Sh RESTORING A PFS FROM A MIRROR 1736A PFS can be restored from a mirror with 1737.Cm mirror-copy . 1738.Cm config 1739data must be copied separately. 1740At last the PFS can be upgraded to master using 1741.Cm pfs-upgrade . 1742.Pp 1743It is not possible to restore the root PFS (PFS# 0) by using mirroring, 1744as the root PFS is always a master PFS. 1745A normal copy (e.g.\& using 1746.Xr cpdup 1 ) 1747must be done, ignoring history. 1748If history is important, old root PFS can me restored to a new PFS, and 1749important directories/files can be 1750.Nm null 1751mounted to the new PFS. 1752.Sh EXIT STATUS 1753.Ex -std 1754.Sh ENVIRONMENT 1755If the following environment variables exist, they will be used by: 1756.Bl -tag -width ".Ev EDITOR" 1757.It Ev EDITOR 1758The editor program specified in the variable 1759.Ev EDITOR 1760will be invoked instead of the default editor, which is 1761.Xr vi 1 . 1762.It Ev VISUAL 1763Same effect as 1764.Ev EDITOR 1765variable. 1766.El 1767.Sh FILES 1768.Bl -tag -width ".It Pa <fs>/var/slaves/<name>" -compact 1769.It Pa <pfs>/snapshots 1770default per PFS snapshots directory 1771.Nm ( HAMMER 1772VERSION 2-) 1773.It Pa /var/hammer/<pfs> 1774default per PFS snapshots directory (not root) 1775.Nm ( HAMMER 1776VERSION 3+) 1777.It Pa /var/hammer/root 1778default snapshots directory for root directory 1779.Nm ( HAMMER 1780VERSION 3+) 1781.It Pa <snapshots>/config 1782per PFS 1783.Nm 1784.Cm cleanup 1785configuration file 1786.Nm ( HAMMER 1787VERSION 2-) 1788.It Pa <fs>/var/slaves/<name> 1789recommended slave PFS snapshots directory 1790.Nm ( HAMMER 1791VERSION 2-) 1792.It Pa <fs>/pfs 1793recommended PFS directory 1794.El 1795.\".Sh EXAMPLES 1796.Sh SEE ALSO 1797.Xr ssh 1 , 1798.Xr undo 1 , 1799.Xr HAMMER 5 , 1800.Xr periodic.conf 5 , 1801.Xr loader 8 , 1802.Xr mount_hammer 8 , 1803.Xr mount_null 8 , 1804.Xr newfs_hammer 8 , 1805.Xr swapcache 8 , 1806.Xr sysctl 8 1807.Sh HISTORY 1808The 1809.Nm 1810utility first appeared in 1811.Dx 1.11 . 1812.Sh AUTHORS 1813.An Matthew Dillon Aq dillon@backplane.com 1814