2 LFSCK: an online file system checker for Lustre
3 ===============================================
5 LFSCK is an online tool to scan, check and repair a Lustre file system that can
6 be used with a file system that is mounted and in use. It contains three main
7 LFSCK components: OI Scrub (primarily of use for ldiskfs-based backend), Layout
8 LFSCK, and Namespace LFSCK. Each component identifies different types of Lustre
11 LFSCK does not verify the on-disk format and assumes that it is consistent. For
12 ldiskfs-based backend, e2fsck from e2fsprogs should be used to ensure the on
13 disk format is consistent. ZFS is designed to always have a valid on-disk
14 structure and as a result, no 'fsck' is necessary.
18 OI scrub is of primary use for ldiskfs-based targets. It maintains the ldiskfs
19 special OI mapping consistency, reconstructs the OI mapping after the target
20 is restored from file-level backup or is otherwise corrupted, and upgrades
21 (if necessary) the OI mapping when target (MDT/OST) is upgraded from a
26 Layout LFSCK is concerned with consistency between metadata targets (MDTs) and
27 object storage targets (OSTs). It automatically corrects inconsistencies where
32 Namespace LFSCK is concerned with consistency across the Lustre namespace.
33 Namespace LFSCK works transparently across single and multiple MDTs.
35 Quick usage instructions
36 ===============================================
40 If you want all LFSCK checks to be run on all MDTs and OSTs, run on MDT0000:
41 # lctl lfsck_start -M $FSNAME-$TARGETNAME -A -t all -r
43 (FSNAME: the specified file system name created during format, e.g. "testfs".
44 TARGETNAME: the target name in the system, e.g. "MDT0000" or "OST0001".)
46 If you want OI Scrub only on one MDT or OST, use this command on the MDT/OST:
47 # lctl lfsck_start -t scrub -M $FSNAME-$TARGETNAME
49 If you want LFSCK Layout or LFSCK Namespace on the given MDT(s), use:
50 # lctl lfsck_start -t namespace -M $FSNAME-$MDTNAME
52 # lctl lfsck_start -t layout -M $FSNAME-$MDTNAME
54 (MDTNAME: the MDT name in the system, e.g. "MDT0000", "MDT0001".)
56 You can trigger multiple LFSCK components via single LFSCK command:
57 # lctl lfsck_start -t namespace -t layout -M $FSNAME-$MDTNAME
59 For more usage, please run:
62 *** Check the status of LFSCK ***
64 By default LFSCK logs all operations to the Lustre internal debug
65 log, which can be dumped to a file on each server with:
66 # lctl debug_kernel /tmp/debug.lfsck
68 However, since the internal debug log is of limited size, it is
69 possible to dump lfsck logs to the console for capture with syslog.
70 # lctl set_param printk=+lfsck
72 Another option is to dump the LFSCK logs to a file directly from the
73 kernel, which is more efficient than logging to the console if there
74 are lots of repairs needed (e.g. after a filesystem upgrade or if the
75 OI files are lost). The following command should be run on all MDS
76 and OSS nodes to generate a log file (maximum 1024MB in size):
77 # lctl debug_daemon start /tmp/debug.lfsck 1024
79 Each LFSCK component has its own status interface on a given target.
80 It is possible to monitor the LFSCK status on the local node via:
81 # lctl lfsck_query -M $FSNAME-$TARGET
83 It is also possible to get type-specific status, for example on
84 the Namespace LFSCK status on the MDT:
85 # lctl get_param -n mdd.$FSNAME-$MDTNAME.lfsck_namespace
87 Or the Layout LFSCK status on the OST:
88 # lctl get_param -n obdfilter.$FSNAME-$OSTNAME.lfsck_layout
89 NOTE: Layout LFSCK also works on a OST.
91 (OSTNAME: the OST name in the system, e.g. "OST0000", "OST0001".)
93 Or the OI Scrub status on the underlying ldiskfs MDT/OST:
94 # lctl get_param -n osd-ldiskfs.$FSNAME-$TARGETNAME.oi_scrub
96 *** Stop the currently running LFSCK ***
98 Run the command on the given MDT/OST:
99 # lctl lfsck_stop -M $FSNAME-$MDTNAME
101 To stop all LFSCK across the system:
102 # lctl lfsck_stop -M $FSNAME -A
105 LFSCK Features Overview
106 ===============================================
109 * control of scanning rate.
110 * automatic checkpoint recovery of an interrupted scan.
111 * monitoring using proc and lctl interfaces.
112 * maintain OI mapping for ldiskfs-based backend.
113 * reconstruction of the OI mapping after the target (MDT/OST) restored from
115 * generate OI mapping for the MDT upgraded from 1.8.
116 * rebuild OI mapping if some of them lost or crashed.
117 * check and repair kinds of namespace related inconsistent issues:
118 * the FID-in-dirent should be consistent with the FID-in-LMA.
119 * the linkEA should be consistent with related name entries.
120 * Dangling name entry: the name entry exists, but related MDT-object does
122 * Orphan MDT-object: the MDT-object exists, but there is no name entry to
124 * Multiple-referenced name entry: more than one MDT-objects point back to
125 the same name entry, but the name entry only references one of them.
126 * Unmatched name entry and MDT-object pairs: the name entry references the
127 MDT-object that has no linkEA for back-reference or points back to another
128 name entry that does not exist or does not reference the MDT-object.
129 * Unmatched object types: the file type in the name entry does not match the
130 type that is claimed by the MDT-object.
131 * Invalid nlink count: the MDT-object's nlink count does not match the number
132 of name entries that reference such MDT-object.
133 * Invalid name hash for striped directory: the name hash for the name entry
134 on a shard of a striped directory does not match the index stored in the
136 * verify layout consistency between MDT and OST:
137 * MDT-object with dangling reference: the MDT-object1 claims that the
138 OST-object1 is its child stripe, but on the OST, the OST-object1 does not
139 exist, or it is not materialized (so does not recognize the MDT-object1 as
141 * Unmatched referenced MDT-object/OST-object pairs: the MDT-object1 claims
142 that the OST-object1 is its child stripe, but the OST-object1 claims that
143 its parent is the MDT-object2 rather than the MDT-object1. On the MDT,
144 the MDT-object2 does not exist, or not recognize the OST-object1 as its
145 child stripe. An additional case exists where the child index stored in
146 the parent layout information does not match the index information stored
148 * Multiple referenced OST-object: the MDT-object1 claims that the OST-object1
149 is its child stripe, but the OST-object1 claims that its parent is the
150 MDT-object2 rather than the MDT-object1. On the other hand, the MDT-object2
151 recognizes the OST-object1 as its child stripe.
152 * Unreferenced (orphan) OST-object: the OST-object1 claims that the
153 MDT-object1 is its parent, but on the MDT, the MDT-object1 does not exist,
154 or it does not recognize the OST-object1 as its child.
158 ===============================================
160 Information about the currently running LFSCK can be found in the following
161 parameter files on the MDS and OSS nodes, using "lctl get_param":
162 mdd.$FSNAME-$MDTNAME.lfsck_layout
163 mdd.$FSNAME-$MDTNAME.lfsck_namespace
164 obdfilter.$FSNAME-$OSTNAME.lfsck_layout
165 osd-ldiskfs.$FSNAME-$TARGETNAME.oi_scrub
168 LFSCK master slave design
169 ===============================================
171 *** Master Engine ***
173 The LFSCK master engine resides on each MDT, and is implemented as a kernel
174 thread in the LFSCK layer. The master engine is responsible for scanning on the
175 MDTs and also controls slave engines on OSTs. Scanning on both MDTs and OSTs
176 occurs in two stages. First-stage scanning will identify and resolve most of
177 inconsistencies. In the second stage, information from the first stage will be
178 used to resolve a remaining set of inconsistencies that had uncertain
179 resolution after only one scan.
182 1. The master engine is started either by the user space command or an
183 excessive number of inconsistency events are detected (defined by
184 osd-ldiskfs.<fsname>-<targetname>.full_scrub_threshold_rate). On starting, the
185 master engine sends RPCs to other MDTs (when necessary) to start other master
186 engines and to related OSTs to start the slave engines.
188 2. The master engine on the MDS scans the MDT device using namespace iteration
189 (described below). For each striped file, it calls the registered LFSCK process
190 handlers to perform the relevant system consistency checks/repairs, which are
191 enumerated in the 'features' section. All objects on OSTs that are never
192 referenced during this scan (because, for example, they are orphans) are
193 recorded in an OST orphan object index on each OST.
195 3. After the MDT completes first-stage system scanning, the master engine sends
196 RPCs to related LFSCK engines on other targets to notify that the first-stage
197 scanning is complete on this MDT. The MDT waits until related targets have
198 completed the first-stage scanning. At this point, the first stage scanning is
199 complete and the second-stage scanning begins.
203 The LFSCK slave engine resides on each OST and is implemented as a kernel
204 thread in the LFSCK layer. This kernel thread drives the first-stage system
207 1. When the slave engine is triggered by the RPC from the master engine in the
208 first-stage scanning, the OST scans the local OST device to generate the
209 in-memory OST orphan object index.
211 2. When the first-stage scanning (for both MDTs and OSTs) is complete a list of
212 non-referenced OST-objects has been accumulated. Only objects that are not
213 accessed during the first stage scan are regarded as potential orphans.
215 3. In the second-stage scanning, the OSTs work to resolve orphan objects in the
216 file system. The OST orphan object index is used as input to the second stage.
217 For each item in the index, the presence of a parent MDT object is verified.
218 Orphan objects will either be relinked to an existing file if found - or moved
219 into a new file in .lustre/lost+found.
221 If multiple MDTs are present, MDTs will check/repair MDT-OST consistency in
222 parallel. To avoid redundant scans of the OST device the slave engine will not
223 begin second-stage system scans until all the master engines complete the
224 first-stage system scan. For each OST there is a single OST orphan object
225 index, regardless of how many MDTs are in the MDT-OST consistency check/repair.
228 Object traversal design reference
229 ===============================================
231 Objects are traversed by LFSCK with two methods: object-table based iteration
232 and namespace based directory traversal.
234 *** Object-table Based Iteration ***
236 The Object Storage Device (OSD) is the abstract layer above a concrete backend
237 file system (i.e. ldiskfs, ZFS, Btrfs, etc.). Each OSD implementation differs
238 internally to support concrete file systems. The object-table based iteration
239 is implemented inside the OSD. It uses the backend special efficient scanning
240 method, such as linear scanning for ldiskfs backend, to scan the local device.
241 Such iteration is presented via the OSD API as a virtual index that contains
242 all the objects that reside on this target.
244 *** Namespace Based Directory Traversal ***
246 In addition to object-table based iteration, there are directory based items
247 that need scanning for namespace consistency. For example, FID-in-dirent and
248 LinkEA are directory based features.
250 A naive approach to namespace traversal would be to descend recursively from
251 the file system root. However, this approach will typically generate random IO,
252 which for performance reasons should be minimized. In addition, one must
253 consider operations (i.e. rename) taking place within a directory that is
254 currently being scanned. For these reasons a hybrid approach to scanning is
257 1. LFSCK begins object-table based iteration.
259 2. If a directory is discovered then namespace traversal begins. LFSCK reads
260 the entries of the directory to verify and repair filename->FID mappings, but
261 does not descend into sub-directories. LFSCK ignores rename operations during
262 the directory traversal because the subsequent object-table based iteration
263 will guarantee processing of renamed objects. Reading directory blocks is a
264 small fraction of the data needed for the objects they reference. In addition,
265 entries in the directory are typically allocated following the directory
266 object on the disk so for many directories the children objects will already
267 be available because of pre-fetch.
269 3. Process each entry in the directory checking the FID-in-dirent and the FID
270 in the object LMA are consistent. Repair if inconsistent. Check also that the
271 linkEA points back to the parent object. Check also that '.' and '..' entries
272 of the directory itself are consistent.
274 4. Once all directory entries are exhausted, return to object-table based
279 ===============================================
281 source code: lustre/lfsck/*.[ch], lustre/osd-ldiskfs/scrub.c
283 operations manual: https://build.whamcloud.com/job/lustre-manual/lastSuccessfulBuild/artifact/lustre_manual.xhtml#dbdoclet.lfsckadmin
285 useful links: https://www.youtube.com/watch?v=jfLo1eYSh2o
286 http://wiki.lustre.org/images/c/c6/Zhuravlev_LFSCK_LUG-2013.pdf
290 ===============================================
292 FID - File IDentifier. A Lustre file system identifies every file and object
293 with a unique 128-bit ID.
295 FID-in-dirent - FID in Directory Entry. To enhance the performance of readdir,
296 the FID of a file is recorded in its directory name entry.
298 linkEA - Link Extended Attributes. When a file is created or hard-linked the
299 parent directory name and FID are recorded as extended attributes to the file.
301 LMA - Lustre Metadata Attributes. A record of Lustre specific attributes, for
302 example HSM state, self-FID, and so on.
304 OI - Object Index. A table that maps FIDs to inodes. On ldiskfs-based targets,
305 this table must be regenerated if a file level restore is performed as inodes
308 OSD - Object storage device. A generic term for a storage device with an
309 interface that extends beyond a block-oriented device interface.