1 <?xml version="1.0" encoding="utf-8"?>
2 <chapter xmlns="http://docbook.org/ns/docbook"
3 xmlns:xl="http://www.w3.org/1999/xlink" version="5.0" xml:lang="en-US"
4 xmlns:xi="http://www.w3.org/2001/XInclude" xml:id="userutilities">
5 <title xml:id="userutilities.title">User Utilities</title>
6 <para>This chapter describes user utilities.</para>
7 <section xml:id="dbdoclet.50438206_94597">
10 <primary>lfs</primary>
12 <literal>lfs</literal>
15 <literal>lfs</literal> utility can be used for user configuration routines
16 and monitoring.</para>
18 <title>Synopsis</title>
21 lfs changelog [--follow] <replaceable>mdt_name</replaceable> [startrec [endrec]]
22 lfs changelog_clear <replaceable>mdt_name id endrec</replaceable>
23 lfs check <replaceable>mds|osts|servers</replaceable>
24 lfs data_version [-nrw] <replaceable>filename</replaceable>
25 lfs df [-i] [-h] [--pool]-p <replaceable>fsname</replaceable>[.<replaceable>pool</replaceable>] [<replaceable>path</replaceable>] [--lazy]
26 lfs find [[!] --atime|-A [-+]N] [[!] --mtime|-M [-+]N]
27 [[!] --ctime|-C [-+]N] [--maxdepth|-D N] [--name|-n <replaceable>pattern</replaceable>]
28 [--print|-p] [--print0|-P] [[!] --obd|-O <replaceable>ost_name</replaceable>[,<replaceable>ost_name...</replaceable>]]
29 [[!] --size|-S [+-]N[kMGTPE]] --type |-t {bcdflpsD}]
30 [[!] --gid|-g|--group|-G <replaceable>gname|gid</replaceable>]
31 [[!] --uid|-u|--user|-U <replaceable>uname|uid</replaceable>]
32 <replaceable>dirname|filename</replaceable>
33 lfs getname [-h]|[path...]
34 lfs getstripe [--obd|-O <replaceable>ost_name</replaceable>] [--quiet|-q] [--verbose|-v]
35 [--stripe-count|-c] [--stripe-index|-i]
36 [--stripe-size|-s] [--pool|-p] [--directory|-d]
37 [--mdt-index|-M] [--recursive|-r] [--raw|-R]
39 <replaceable>dirname|filename</replaceable> ...
40 lfs setstripe [--size|-s stripe_size] [--stripe-count|-c <replaceable>stripe_count</replaceable>]
41 [--overstripe-count|-C <replaceable>stripe_count</replaceable>]
42 [--stripe-index|-i <replaceable>start_ost_index</replaceable>]
43 [--ost-list|-o <replaceable>ost_indicies</replaceable>]
44 [--pool|-p <replaceable>pool</replaceable>]
45 <replaceable>dirname|filename</replaceable>
46 lfs setstripe -d <replaceable>dir</replaceable>
48 lfs pool_list <replaceable>filesystem</replaceable>[.<replaceable>pool</replaceable>]| <replaceable>pathname</replaceable>
49 lfs quota [-q] [-v] [-h] [-o <replaceable>obd_uuid</replaceable>|-I <replaceable>ost_idx</replaceable>|-i <replaceable>mdt_idx</replaceable>]
50 [-u <replaceable>username|uid|-g</replaceable> <replaceable>group|gid</replaceable>|-p <replaceable>projid</replaceable>] <replaceable>/mount_point</replaceable>
51 lfs quota -t -u|-g|-p <replaceable>/mount_point</replaceable>
52 lfs setquota {-u|--user|-g|--group|-p|--project} <replaceable>uname|uid|gname|gid|projid</replaceable>
53 [--block-softlimit <replaceable>block_softlimit</replaceable>]
54 [--block-hardlimit <replaceable>block_hardlimit</replaceable>]
55 [--inode-softlimit <replaceable>inode_softlimit</replaceable>]
56 [--inode-hardlimit <replaceable>inode_hardlimit</replaceable>]
57 <replaceable>/mount_point</replaceable>
58 lfs setquota -u|--user|-g|--group|-p|--project <replaceable>uname|uid|gname|gid|projid</replaceable>
59 [-b <replaceable>block_softlimit</replaceable>] [-B <replaceable>block_hardlimit</replaceable>]
60 [-i <replaceable>inode-softlimit</replaceable>] [-I <replaceable>inode_hardlimit</replaceable>]
61 <replaceable>/mount_point</replaceable>
62 lfs setquota -t -u|-g|-p [--block-grace <replaceable>block_grace</replaceable>]
63 [--inode-grace <replaceable>inode_grace</replaceable>]
64 <replaceable>/mount_point</replaceable>
65 lfs setquota -t -u|-g|-p [-b <replaceable>block_grace</replaceable>] [-i <replaceable>inode_grace</replaceable>]
66 <replaceable>/mount_point</replaceable>
70 <para>In the above example, the
72 <replaceable>/mount_point</replaceable>
73 </literal> parameter refers to the mount point of the Lustre file
77 <para>The old lfs quota output was very detailed and contained
78 cluster-wide quota statistics (including cluster-wide limits for a
79 user/group and cluster-wide usage for a user/group), as well as
80 statistics for each MDS/OST. Now,
81 <literal>lfs quota</literal> has been updated to provide only
82 cluster-wide statistics, by default. To obtain the full report of
83 cluster-wide limits, usage and statistics, use the
84 <literal>-v</literal> option with
85 <literal>lfs quota</literal>.</para>
87 <para condition="l24">
88 The <literal>quotacheck</literal>, <literal>quotaon</literal> and <literal>quotaoff</literal>
89 sub-commands were deprecated in the Lustre 2.4 release, and removed completely in
90 the Lustre 2.8 release. See <xref linkend="enabling_disk_quotas"/> for details on
91 configuring and checking quotas.
95 <title>Description</title>
97 <literal>lfs</literal> utility is used to create a new file with a
98 specific striping pattern, determine the default striping pattern, gather
99 the extended attributes (object numbers and location) for a specific
100 file, find files with specific attributes, list OST information or set
101 quota limits. It can be invoked interactively without any arguments or in
102 a non-interactive mode with one of the supported arguments.</para>
105 <title>Options</title>
107 <literal>lfs</literal> options are listed and described below. For a
108 complete list of available options, type help at the
109 <literal>lfs</literal> prompt.</para>
110 <informaltable frame="all">
112 <colspec colname="c1" colwidth="20*" />
113 <colspec colname="c2" colwidth="30*" />
114 <colspec colname="c3" colwidth="50*" />
117 <entry nameend="c2" namest="c1">
119 <emphasis role="bold">Option</emphasis>
124 <emphasis role="bold">Description</emphasis>
131 <entry nameend="c2" namest="c1">
133 <literal>changelog</literal>
137 <para>Shows the metadata changes on an MDT. Start and end
138 points are optional. The
139 <literal>--follow</literal> option blocks on new changes; this
140 option is only valid when run directly on the MDT node.</para>
144 <entry nameend="c2" namest="c1">
146 <literal>changelog_clear</literal>
150 <para>Indicates that changelog records previous to
152 <replaceable>endrec</replaceable>
153 </literal> are no longer of interest to a particular consumer
155 <replaceable>id</replaceable>
156 </literal>, potentially allowing the MDT to free up disk space.
159 <replaceable>endrec</replaceable>
160 </literal> of 0 indicates the current last record. Changelog
161 consumers must be registered on the MDT node using
162 <literal>lctl</literal>.</para>
166 <entry nameend="c2" namest="c1">
167 <literal>check</literal>
170 <para>Displays the status of MDS or OSTs (as specified in the
171 command) or all servers (MDS and OSTs).</para>
175 <entry nameend="c2" namest="c1">
177 <literal>data_version [-nrw]
178 <replaceable>filename</replaceable></literal>
182 <para>Displays the current version of file data. If
183 <literal>-n</literal> is specified, the data version is read
184 without taking a lock. As a consequence, the data version could
185 be outdated if there are dirty caches on filesystem clients, but
186 this option will not force data flushes and has less of an
187 impact on the filesystem. If <literal>-r</literal> is specified,
188 the data version is read after dirty pages on clients are
189 flushed. If <literal>-w</literal> is specified, the data version
190 is read after all caching pages on clients are flushed.
193 Even with <literal>-r</literal> or <literal>-w</literal>, race
194 conditions are possible and the data version should be checked
195 before and after an operation to be confident the data did not
199 The data version is the sum of the last committed transaction
200 numbers of all data objects of a file. It is used by HSM policy
201 engines for verifying that file data has not been changed during
202 an archive operation or before a release operation, and by OST
203 migration, primarily for verifying that file data has not been
204 changed during a data copy, when done in non-blocking mode.
209 <entry nameend="c2" namest="c1">
210 <literal>df [-i] [-h] [--pool|-p
211 <replaceable>fsname</replaceable>[.
212 <replaceable>pool</replaceable>] [
213 <replaceable>path</replaceable>] [--lazy]</literal>
217 <literal>-i</literal> to report file system disk space usage or
218 inode usage of each MDT or OST or, if a pool is specified with
220 <literal>-p</literal> option, a subset of OSTs.</para>
221 <para>By default, the usage of all mounted Lustre file systems
223 <literal>path</literal> option is included, only the usage for
224 the specified file system is reported. If the
225 <literal>-h</literal> option is included, the output is printed
226 in human-readable format, using SI base-2 suffixes for
227 <emphasis role="bold">M</emphasis>ega-,
228 <emphasis role="bold">G</emphasis>iga-,
229 <emphasis role="bold">T</emphasis>era-,
230 <emphasis role="bold">P</emphasis>eta-, or
231 <emphasis role="bold">E</emphasis>xabytes.</para>
233 <literal>--lazy</literal> option is specified, any OST that is
234 currently disconnected from the client will be skipped. Using
236 <literal>--lazy</literal> option prevents the
237 <literal>df</literal> output from being blocked when an OST is
238 offline. Only the space on the OSTs that can currently be
239 accessed are returned. The
240 <literal>llite.*.lazystatfs</literal> tunable can be enabled to
241 make this the default behaviour for all
242 <literal>statfs()</literal> operations.</para>
246 <entry nameend="c2" namest="c1">
248 <literal>find</literal>
252 <para>Searches the directory tree rooted at the given
253 directory/filename for files that match the given
256 <literal>!</literal> before an option negates its meaning (files
257 NOT matching the parameter). Using
258 <literal>+</literal> before a numeric value means files with the
259 parameter OR MORE. Using
260 <literal>-</literal> before a numeric value means files with the
261 parameter OR LESS.</para>
267 <literal>--atime</literal>
270 <para>File was last accessed N*24 hours ago. (There is no
272 <literal>atime</literal> is kept coherent across the
274 <para>OSTs store a transient
275 <literal>atime</literal> that is updated when clients do read
277 <literal>atime</literal> is written to the MDS when the file is
278 closed. However, on-disk atime is only updated if it is more
279 than 60 seconds old (
280 <literal>/proc/fs/lustre/mds/*/max_atime_diff</literal>). The
281 Lustre software considers the latest
282 <literal>atime</literal> from all OSTs. If a
283 <literal>setattr</literal> is set by user, then it is updated on
284 both the MDS and OST, allowing the
285 <literal>atime</literal> to go backward.</para>
294 <literal>--ctime</literal>
298 <para>File status was last changed N*24 hours ago.</para>
307 <literal>--mtime</literal>
311 <para>File data was last modified N*24 hours ago.</para>
320 <literal>--obd</literal>
324 <para>File has an object on a specific OST(s).</para>
333 <literal>--size</literal>
337 <para>File has a size in bytes, or kilo-, Mega-, Giga-, Tera-,
338 Peta- or Exabytes if a suffix is given.</para>
347 <literal>--type</literal>
351 <para>File has the type - block, character, directory, pipe,
352 file, symlink, socket or door (used in Solaris operating
362 <literal>--uid</literal>
366 <para>File has a specific numeric user ID.</para>
375 <literal>--user</literal>
379 <para>File owned by a specific user (numeric user ID
389 <literal>--gid</literal>
393 <para>File has a specific group ID.</para>
402 <literal>--group</literal>
406 <para>File belongs to a specific group (numeric group ID
416 <literal>-maxdepth</literal></para>
419 <para>Limits find to descend at most N levels of the directory
429 <literal>--print</literal>/
430 <literal>--print0</literal></para>
433 <para>Prints the full filename, followed by a new line or NULL
434 character correspondingly.</para>
438 <entry nameend="c2" namest="c1">
440 <literal>osts [path]</literal>
444 <para>Lists all OSTs for the file system. If a path located on
445 a mounted Lustre file system is specified, then only OSTs
446 belonging to this file system are displayed.</para>
450 <entry nameend="c2" namest="c1">
452 <literal>getname [path...]</literal>
456 <para>List each Lustre file system instance associated with
457 each Lustre mount point. If no path is specified, all Lustre
458 mount points are interrogated. If a list of paths is provided,
459 the instance of each path is provided. If the path is not a
460 Lustre instance 'No such device' is returned.</para>
464 <entry nameend="c2" namest="c1">
466 <literal>getstripe</literal>
470 <para>Lists striping information for a given filename or
471 directory. By default, the stripe count, stripe size and offset
473 <para>If you only want specific striping information, then the
475 <literal>--stripe-count</literal>,
476 <literal>--stripe-size</literal>,
477 <literal>--stripe-index</literal>,
478 <literal>--layout</literal>, or
479 <literal>--pool</literal> plus various combinations of these
480 options can be used to retrieve specific information.</para>
482 <literal>--raw</literal> option is specified, the stripe
483 information is printed without substituting the file system
484 default values for unspecified fields. If the striping EA is
485 not set, 0, 0, and -1 will be printed for the stripe count,
486 size, and offset respectively.</para>
487 <para>The <literal>--mdt-index</literal> prints the index of
488 the MDT for a given directory. See
489 <xref linkend="lustremaint.rmremotedir" />.</para>
499 <replaceable>ost_name</replaceable></literal>
503 <para>Lists files that have an object on a specific OST.</para>
512 <literal>--quiet</literal>
516 <para>Lists details about the file's object ID
526 <literal>--verbose</literal>
530 <para>Prints additional striping information.</para>
539 <literal>--stripe-count</literal>
543 <para>Lists the stripe count (how many OSTs to use).</para>
552 <literal>--index</literal>
556 <para>Lists the index for each OST in the file system.</para>
565 <literal>--offset</literal>
569 <para>Lists the OST index on which file striping starts.</para>
578 <literal>--pool</literal>
582 <para>Lists the pools to which a file belongs.</para>
591 <literal>--size</literal>
595 <para>Lists the stripe size (how much data to write to one OST
596 before moving to the next OST).</para>
605 <literal>--directory</literal>
609 <para>Lists entries about a specified directory instead of its
610 contents (in the same manner as
611 <literal>ls -d</literal>).</para>
620 <literal>--recursive</literal>
624 <para>Recurses into all sub-directories.</para>
628 <entry nameend="c2" namest="c1">
630 <literal>setstripe</literal>
634 <para>Create new files with a specific file layout (stripe
635 pattern) configuration.
637 <para>The file cannot exist prior to using
638 <literal>setstripe</literal>. A directory must exist prior to
640 <literal>setstripe</literal>.</para>
650 <literal>--stripe-count stripe_cnt</literal>
654 <para>Number of OSTs over which to stripe a file. A
655 <literal>stripe_cnt</literal> of 0 uses the file system-wide
656 default stripe count (default is 1). A
657 <literal>stripe_cnt</literal> of -1 stripes over all available
667 <literal>--overstripe-count stripe_cnt</literal>
671 <para>The same as --stripe-count, but allows overstriping,
672 which will place more than one stripe per OST if
673 <literal>stripe_cnt</literal> is greater than the number of
674 OSTs. Overstriping is useful for matching the number of stripes
675 to the number of processes, or with very fast OSTs, where one
676 stripe per OST is not enough to get full performance.</para>
685 <literal>--size stripe_size</literal>
687 <para>The default stripe-size is 0. The default start-ost is
688 -1. Do NOT confuse them! If you set start-ost to 0, all new
689 file creations occur on OST 0 (seldom a good idea).</para>
690 </footnote> </para>
693 <para>Number of bytes to store on an OST before moving to the
694 next OST. A stripe_size of 0 uses the file system's default
695 stripe size, (default is 1 MB). Can be specified with
696 <emphasis role="bold">k</emphasis>(KB),
697 <emphasis role="bold">m</emphasis>(MB), or
698 <emphasis role="bold">g</emphasis>(GB), respectively.</para>
707 <literal>--stripe-index start_ost_index</literal>
711 <para>The OST index (base 10, starting at 0) on which to start
712 striping for this file. A start_ost_index value of -1 allows
713 the MDS to choose the starting index. This is the default
714 value, and it means that the MDS selects the starting OST as it
715 wants. We strongly recommend selecting this default, as it
716 allows space and load balancing to be done by the MDS as
718 <literal>start_ost_index</literal> value has no relevance on
719 whether the MDS will use round-robin or QoS weighted allocation
720 for the remaining stripes in the file.</para>
729 <literal>--ost-index ost_indices</literal>
733 <para>This option is used to specify the exact stripe
734 layout on the the file system. <literal>ost_indices</literal>
735 is a list of OSTs referenced by their indices and index ranges
736 separated by commas, e.g. <literal>1,2-4,7</literal>.</para>
746 <replaceable>pool</replaceable></literal>
750 <para>Name of the pre-defined pool of OSTs (see
751 <xref linkend="dbdoclet.50438219_38274" />) that will be used
753 <literal>stripe_cnt</literal>,
754 <literal>stripe_size</literal> and
755 <literal>start_ost</literal> values are used as well. The
756 start-ost value must be part of the pool or an error is
761 <entry nameend="c2" namest="c1">
763 <literal>setstripe -d</literal>
767 <para>Deletes default striping on the specified
772 <entry nameend="c2" namest="c1">
774 <literal>pool_list {filesystem}[.poolname]|{pathname}</literal>
778 <para>Lists pools in the file system or pathname, or OSTs in
779 the file system's pool.</para>
783 <entry nameend="c2" namest="c1">
785 <literal>quota [-q] [-v] [-o
786 <replaceable>obd_uuid</replaceable>|-i
787 <replaceable>mdt_idx</replaceable>|-I
788 <replaceable>ost_idx</replaceable>] [-u|-g|-p
789 <replaceable>uname|uid|gname|gid|projid]</replaceable>
790 <replaceable>/mount_point</replaceable></literal>
795 <para>Displays disk usage and limits, either for the full file
796 system or for objects on a specific OBD. A user or group name
797 or an usr, group and project ID can be specified. If all user,
798 group project ID are omitted, quotas for the current UID/GID
799 are shown. The <literal>-q</literal> option disables printing
800 of additional descriptions (including column titles). It fills
801 in blank spaces in the
802 <literal>grace</literal> column with zeros (when there is no
803 grace period set), to ensure that the number of columns is
805 <literal>-v</literal> option provides more verbose (per-OBD
806 statistics) output.</para>
810 <entry nameend="c2" namest="c1">
813 <replaceable>-u|-g|-p</replaceable>
814 <replaceable>/mount_point</replaceable></literal>
818 <para>Displays block and inode grace times for user (
819 <literal>-u</literal>) or group (
820 <literal>-g</literal>) or project (
821 <literal>-p</literal>) quotas.</para>
825 <entry nameend="c2" namest="c1">
827 <literal>setquota {-u|-g|-p
828 <replaceable>uname|uid|gname|gid|projid}</replaceable>
830 <replaceable>block_softlimit</replaceable>]
832 <replaceable>block_hardlimit</replaceable>]
834 <replaceable>inode_softlimit</replaceable>]
836 <replaceable>inode_hardlimit</replaceable>]
837 <replaceable>/mount_point</replaceable></literal>
841 <para>Sets file system quotas for users, groups or one project.
842 Limits can be specified with
843 <literal>--{block|inode}-{softlimit|hardlimit}</literal> or
844 their short equivalents
845 <literal>-b</literal>,
846 <literal>-B</literal>,
847 <literal>-i</literal>,
848 <literal>-I</literal>. Users can set 1, 2, 3 or 4 limits.
851 <literal>setquota</literal> interface is supported, but it may
852 be removed in a future Lustre software release.</para>
853 </footnote>Also, limits can be specified with special suffixes,
854 -b, -k, -m, -g, -t, and -p to indicate units of 1, 2^10, 2^20,
855 2^30, 2^40 and 2^50, respectively. By default, the block limits
856 unit is 1 kilobyte (1,024), and block limits are always
857 kilobyte-grained (even if specified in bytes). See
858 <xref linkend="dbdoclet.50438206_11903" />.</para>
862 <entry nameend="c2" namest="c1">
864 <literal>setquota -t -u|-g|-p [--block-grace
865 <replaceable>block_grace</replaceable>] [--inode-grace
866 <replaceable>inode_grace</replaceable>]
867 <replaceable>/mount_point</replaceable></literal>
871 <para>Sets the file system quota grace times for users or
872 groups. Grace time is specified in '
873 <literal>XXwXXdXXhXXmXXs</literal>' format or as an integer
875 <xref linkend="dbdoclet.50438206_11903" />.</para>
879 <entry nameend="c2" namest="c1">
881 <literal>help</literal>
885 <para>Provides brief help on various
886 <literal>lfs</literal> arguments.</para>
890 <entry nameend="c2" namest="c1">
892 <literal>exit/quit</literal>
896 <para>Quits the interactive
897 <literal>lfs</literal> session.</para>
904 <section remap="h5" xml:id="dbdoclet.50438206_11903">
905 <title>Examples</title>
906 <para>Creates a file striped on two OSTs with 128 KB on each
909 $ lfs setstripe -s 128k -c 2 /mnt/lustre/file1
911 <para>Deletes a default stripe pattern on a given directory. New files
912 use the default striping pattern.</para>
914 $ lfs setstripe -d /mnt/lustre/dir
916 <para>Lists the detailed object allocation of a given file.</para>
918 $ lfs getstripe -v /mnt/lustre/file1
920 <para>List all the mounted Lustre file systems and corresponding Lustre
925 <para>Efficiently lists all files in a given directory and its
926 subdirectories.</para>
928 $ lfs find /mnt/lustre
930 <para>Recursively lists all regular files in a given directory more than
933 $ lfs find /mnt/lustre -mtime +30 -type f -print
935 <para>Recursively lists all files in a given directory that have objects
936 on OST2-UUID. The lfs check servers command checks the status of all
937 servers (MDT and OSTs).</para>
939 $ lfs find --obd OST2-UUID /mnt/lustre/
941 <para>Lists all OSTs in the file system.</para>
945 <para>Lists space usage per OST and MDT in human-readable format.</para>
949 <para>Lists inode usage per OST and MDT.</para>
953 <para>List space or inode usage for a specific OST pool.</para>
956 <replaceable>filesystem</replaceable>[.
957 <replaceable>pool</replaceable>] |
958 <replaceable>pathname</replaceable>
960 <para>List quotas of user 'bob'.</para>
962 $ lfs quota -u bob /mnt/lustre
964 <para>List quotas of project ID '1'.</para>
966 $ lfs quota -p 1 /mnt/lustre
968 <para>Show grace times for user quotas on
969 <literal>/mnt/lustre</literal>.</para>
971 $ lfs quota -t -u /mnt/lustre
973 <para>Sets quotas of user 'bob', with a 1 GB block quota hardlimit and a
974 2 GB block quota softlimit.</para>
976 $ lfs setquota -u bob --block-softlimit 2000000 --block-hardlimit 1000000
979 <para>Sets grace times for user quotas: 1000 seconds for block quotas, 1
980 week and 4 days for inode quotas.</para>
982 $ lfs setquota -t -u --block-grace 1000 --inode-grace 1w4d /mnt/lustre
984 <para>Checks the status of all servers (MDT, OST)</para>
988 <para>Creates a file striped on two OSTs from the pool
989 <literal>my_pool</literal></para>
991 $ lfs setstripe --pool my_pool -c 2 /mnt/lustre/file
993 <para>Lists the pools defined for the mounted Lustre file system
994 <literal>/mnt/lustre</literal></para>
996 $ lfs pool_list /mnt/lustre/
998 <para>Lists the OSTs which are members of the pool
999 <literal>my_pool</literal> in file system
1000 <literal>my_fs</literal></para>
1002 $ lfs pool_list my_fs.my_pool
1004 <para>Finds all directories/files associated with
1005 <literal>poolA</literal>.</para>
1007 $ lfs find /mnt/lustre --pool poolA
1009 <para>Finds all directories/files not associated with a pool.</para>
1011 $ lfs find /mnt//lustre --pool ""
1013 <para>Finds all directories/files associated with pool.</para>
1015 $ lfs find /mnt/lustre ! --pool ""
1017 <para>Associates a directory with the pool
1018 <literal>my_pool</literal>, so all new files and directories are created
1021 $ lfs setstripe --pool my_pool /mnt/lustre/dir
1024 <section remap="h5">
1025 <title>See Also</title>
1027 <xref linkend="dbdoclet.50438219_38274" />
1031 <section xml:id="dbdoclet.lfs_migrate">
1034 <primary>lfs_migrate</primary>
1036 <literal>lfs_migrate</literal>
1039 <literal>lfs_migrate</literal> utility is a simple to migrate file
1040 <emphasis>data</emphasis> between OSTs.</para>
1041 <section remap="h5">
1042 <title>Synopsis</title>
1044 lfs_migrate [<replaceable>lfs_setstripe_options</replaceable>]
1045 [-h] [-n] [-q] [-R] [-s] [-y] [-0] [file|directory ...]
1048 <section remap="h5">
1049 <title>Description</title>
1051 <literal>lfs_migrate</literal> utility is a tool to assist migration
1052 of file data between Lustre OSTs. The utility copies each specified
1053 file to a temporary file using supplied <literal>lfs setstripe</literal>
1054 options, if any, optionally verifies the file contents have not changed,
1055 and then swaps the layout (OST objects) from the temporary file and the
1056 original file (for Lustre 2.5 and later), or renames the temporary file
1057 to the original filename. This allows the user/administrator to balance
1058 space usage between OSTs, or move files off OSTs that are starting to show
1059 hardware problems (though are still functional) or will be removed.</para>
1061 <para>For versions of Lustre before 2.5,
1062 <literal>lfs_migrate</literal> was not integrated with the MDS at all.
1063 That made it UNSAFE for use on files that were being modified by other
1064 applications, since the file was migrated through a copy and rename of
1065 the file. With Lustre 2.5 and later, the new file layout is swapped
1066 with the existing file layout, which ensures that the user-visible
1067 inode number is kept, and open file handles and locks on the file are
1070 <para>Files to be migrated can be specified as command-line arguments. If
1071 a directory is specified on the command-line then all files within the
1072 directory are migrated. If no files are specified on the command-line,
1073 then a list of files is read from the standard input, making
1074 <literal>lfs_migrate</literal> suitable for use with
1075 <literal>lfs find</literal> to locate files on specific OSTs and/or
1076 matching other file attributes, and other tools that generate a list
1077 of files on standard output.</para>
1078 <para>Unless otherwise specified through command-line options, the
1079 file allocation policies on the MDS dictate where the new files
1080 are placed, taking into account whether specific OSTs have been
1081 disabled on the MDS via <literal>lctl</literal> (preventing new
1082 files from being allocated there), whether some OSTs are overly full
1083 (reducing the number of files placed on those OSTs), or if there is
1084 a specific default file striping for the parent directory (potentially
1085 changing the stripe count, stripe size, OST pool, or OST index of a
1089 <literal>lfs_migrate</literal> utility can also be used in some cases to
1092 <primary>fragmentation</primary>
1093 </indexterm>fragmentation. File fragmentation will typically reduce
1094 Lustre file system performance. File fragmentation may be observed on
1095 an aged file system and will commonly occur if the file was written by
1096 many threads. Provided there is sufficient free space (or if it was
1097 written when the file system was nearly full) that is less fragmented
1098 than the file being copied, re-writing a file with
1099 <literal>lfs_migrate</literal> will result in a migrated file with
1100 reduced fragmentation. The tool
1101 <literal>filefrag</literal> can be used to report file fragmentation.
1103 <xref linkend="dbdoclet.50438206_75125" /></para>
1106 <para>As long as a file has extent lengths of tens of megabytes (
1107 <replaceable>read_bandwidth * seek_time</replaceable>) or more, the
1108 read performance for the file will not be significantly impacted by
1109 fragmentation, since the read pipeline can be filled by large reads
1110 from disk even with an occasional disk seek.</para>
1113 <section remap="h5">
1114 <title>Options</title>
1115 <para>Options supporting
1116 <literal>lfs_migrate</literal> are described below.</para>
1117 <informaltable frame="all">
1119 <colspec colname="c1" colwidth="50*" />
1120 <colspec colname="c2" colwidth="50*" />
1125 <emphasis role="bold">Option</emphasis>
1130 <emphasis role="bold">Description</emphasis>
1140 <replaceable>stripecount</replaceable></literal>
1144 <para>Restripe file using the specified stripe count. This
1145 option may not be specified at the same time as the
1146 <literal>-R</literal> option.</para>
1152 <literal>-h</literal>
1156 <para>Display help information.</para>
1161 <literal>-l</literal>
1164 <para>Migrate files with hard links (skips, by default). Files
1165 with multiple hard links are split into multiple separate files
1167 <literal>lfs_migrate</literal>, so they are skipped, by
1168 default, to avoid breaking the hard links.</para>
1173 <literal>-n</literal>
1176 <para>Only print the names of files to be migrated.</para>
1182 <literal>-q</literal>
1186 <para>Run quietly (does not print filenames or status).</para>
1191 <literal>-R</literal>
1193 <entry>Restripe file using default directory striping instead of
1194 keeping striping. This option may not be specified at the same
1196 <literal>-c</literal> option.</entry>
1200 <literal>-s</literal>
1202 <entry>Skip file data comparison after migrate. Default is to
1203 compare migrated file against original to verify
1204 correctness.</entry>
1209 <literal>-y</literal>
1214 <literal>y</literal>' to usage warning without prompting (for
1215 scripts, use with caution).</para>
1220 <literal>-0</literal>
1222 <entry>Expect NUL-terminated filenames on standard input, as
1223 generated by <literal>lfs find -print0</literal> or
1224 <literal>find -print0</literal>. This allows filenames with
1225 embedded newlines to be handled correctly.
1232 <section remap="h5">
1233 <title>Examples</title>
1234 <para>Rebalance all files in
1235 <literal>/mnt/lustre/dir</literal>:</para>
1237 $ lfs_migrate /mnt/lustre/dir
1239 <para>Migrate files in /test filesystem on OST0004 larger than 4 GB in
1240 size and older than a day old:</para>
1242 $ lfs find /test -obd test-OST0004 -size +4G -mtime +1 | lfs_migrate -y
1245 <section remap="h5">
1246 <title>See Also</title>
1248 <xref linkend="dbdoclet.50438206_94597" />
1252 <section xml:id="dbdoclet.50438206_75125">
1255 <primary>filefrag</primary>
1257 <literal>filefrag</literal>
1260 <literal>e2fsprogs</literal> package contains the
1261 <literal>filefrag</literal> tool which reports the extent of file
1262 fragmentation.</para>
1263 <section remap="h5">
1264 <title>Synopsis</title>
1266 filefrag [ -belsv ] [ files... ]
1269 <section remap="h5">
1270 <title>Description</title>
1272 <literal>filefrag</literal> utility reports the extent of fragmentation in
1274 <literal>filefrag</literal> utility obtains the extent information from
1275 Lustre files using the
1276 <literal>FIEMAP ioctl</literal>, which is efficient and fast, even for
1277 very large files.</para>
1278 <para>In default mode
1280 <para>The default mode is faster than the verbose/extent mode since it
1281 only counts the number of extents.</para>
1283 <literal>filefrag</literal> prints the number of physically discontiguous
1284 extents in the file. In extent or verbose mode, each extent is printed
1285 with details such as the blocks allocated on each OST. For a Lustre file
1286 system, the extents are printed in device offset order (i.e. all of the
1287 extents for one OST first, then the next OST, etc.), not file logical
1288 offset order. If the file logical offset order was used, the Lustre
1289 striping would make the output very verbose and difficult to see if there
1290 was file fragmentation or not.</para>
1292 <para>Note that as long as a file has extent lengths of tens of
1293 megabytes or more (i.e.
1294 <replaceable>read_bandwidth * seek_time >
1295 extent_length</replaceable>), the read performance for the file will
1296 not be significantly impacted by fragmentation, since file readahead
1297 can fully utilize the disk disk bandwidth even with occasional
1300 <para>In default mode
1302 <para>The default mode is faster than the verbose/extent mode.</para>
1304 <literal>filefrag</literal> returns the number of physically discontiguous
1305 extents in the file. In extent or verbose mode, each extent is printed
1306 with details. For a Lustre file system, the extents are printed in device
1307 offset order, not logical offset order.</para>
1309 <section remap="h5">
1310 <title>Options</title>
1311 <para>The options and descriptions for the
1312 <literal>filefrag</literal> utility are listed below.</para>
1313 <informaltable frame="all">
1315 <colspec colname="c1" colwidth="50*" />
1316 <colspec colname="c2" colwidth="50*" />
1321 <emphasis role="bold">Option</emphasis>
1326 <emphasis role="bold">Description</emphasis>
1335 <literal>-b</literal>
1339 <para>Uses the 1024-byte blocksize for the output. By default,
1340 this blocksize is used by the Lustre file system, since OSTs
1341 may use different block sizes.</para>
1347 <literal>-e</literal>
1351 <para>Uses the extent mode when printing the output. This is
1352 the default for Lustre files in verbose mode.</para>
1358 <literal>-l</literal>
1362 <para>Displays extents in LUN offset order. This is the only
1363 available mode for Lustre.</para>
1369 <literal>-s</literal>
1373 <para>Synchronizes any unwritten file data to disk before
1374 requesting the mapping.</para>
1380 <literal>-v</literal>
1384 <para>Prints the file's layout in verbose mode when checking
1385 file fragmentation, including the logical to physical mapping
1386 for each extent in the file and the OST index.</para>
1393 <section remap="h5">
1394 <title>Examples</title>
1395 <para>Lists default output.</para>
1397 $ filefrag /mnt/lustre/foo
1398 /mnt/lustre/foo: 13 extents found
1400 <para>Lists verbose output in extent format.</para>
1402 $ filefrag -v /mnt/lustre/foo
1403 Filesystem type is: bd00bd0
1404 File size of /mnt/lustre/foo is 1468297786 (1433888 blocks of 1024 bytes)
1405 ext: device_logical: physical_offset: length: dev: flags:
1406 0: 0.. 122879: 2804679680..2804802559: 122880: 0002: network
1407 1: 122880.. 245759: 2804817920..2804940799: 122880: 0002: network
1408 2: 245760.. 278527: 2804948992..2804981759: 32768: 0002: network
1409 3: 278528.. 360447: 2804982784..2805064703: 81920: 0002: network
1410 4: 360448.. 483327: 2805080064..2805202943: 122880: 0002: network
1411 5: 483328.. 606207: 2805211136..2805334015: 122880: 0002: network
1412 6: 606208.. 729087: 2805342208..2805465087: 122880: 0002: network
1413 7: 729088.. 851967: 2805473280..2805596159: 122880: 0002: network
1414 8: 851968.. 974847: 2805604352..2805727231: 122880: 0002: network
1415 9: 974848.. 1097727: 2805735424..2805858303: 122880: 0002: network
1416 10: 1097728.. 1220607: 2805866496..2805989375: 122880: 0002: network
1417 11: 1220608.. 1343487: 2805997568..2806120447: 122880: 0002: network
1418 12: 1343488.. 1433599: 2806128640..2806218751: 90112: 0002: network
1419 /mnt/lustre/foo: 13 extents found
1423 <section xml:id="dbdoclet.50438206_86244">
1426 <primary>mount</primary>
1428 <literal>mount</literal>
1431 <literal>mount(8)</literal> Linux command is used to mount a Lustre file
1432 system. When mounting a Lustre file system, mount(8) executes the
1433 <literal>/sbin/mount.lustre</literal> command to complete the mount. The
1434 mount command supports these options specific to a Lustre file
1436 <informaltable frame="all">
1438 <colspec colname="c1" colwidth="50*" />
1439 <colspec colname="c2" colwidth="50*" />
1444 <emphasis role="bold">Server options</emphasis>
1449 <emphasis role="bold">Description</emphasis>
1458 <literal>abort_recov</literal>
1462 <para>Aborts recovery when starting a target</para>
1468 <literal>nosvc</literal>
1472 <para>Starts only MGS/MGC servers</para>
1478 <literal>nomgs</literal>
1482 <para>Start a MDT with a co-located MGS without starting the
1489 <literal>exclude</literal>
1493 <para>Starts with a dead OST</para>
1499 <literal>md_stripe_cache_size</literal>
1503 <para>Sets the stripe cache size for server side disk with a
1504 striped raid configuration</para>
1510 <informaltable frame="all">
1512 <colspec colname="c1" colwidth="50*" />
1513 <colspec colname="c2" colwidth="50*" />
1518 <emphasis role="bold">Client options</emphasis>
1523 <emphasis role="bold">Description</emphasis>
1532 <literal>flock/noflock/localflock</literal>
1536 <para>Enables/disables global flock or local flock support</para>
1542 <literal>user_xattr/nouser_xattr</literal>
1546 <para>Enables/disables user-extended attributes</para>
1552 <literal>user_fid2path/nouser_fid2path</literal>
1556 <para>Enables/disables FID to path translation by
1557 regular users</para>
1563 <literal>retry=</literal>
1567 <para>Number of times a client will retry to mount the file
1575 <section xml:id="dbdoclet.50438206_56217">
1576 <title>Handling Timeouts</title>
1577 <para>Timeouts are the most common cause of hung applications. After a
1578 timeout involving an MDS or failover OST, applications attempting to access
1579 the disconnected resource wait until the connection gets
1581 <para>When a client performs any remote operation, it gives the server a
1582 reasonable amount of time to respond. If a server does not reply either due
1583 to a down network, hung server, or any other reason, a timeout occurs which
1584 requires a recovery.</para>
1585 <para>If a timeout occurs, a message (similar to this one), appears on the
1586 console of the client, and in
1587 <literal>/var/log/messages</literal>:</para>
1589 LustreError: 26597:(client.c:810:ptlrpc_expire_one_request()) @@@ timeout
1591 req@a2d45200 x5886/t0 o38->mds_svc_UUID@NID_mds_UUID:12 lens 168/64 ref 1 fl
1597 <!--vim:expandtab:shiftwidth=2:tabstop=8:-->