2 .\" Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved.
4 .\" Copyright (c) 2012, 2017, Intel Corporation.
6 .\" This file may be copied under the terms of the GNU Public License v2.
8 .TH mount.lustre 8 "2019 Jul 30" Lustre "configuration utilities"
10 mount.lustre \- mount helper for Lustre client
11 mount.lustre_tgt \- mount helper for Lustre target service
14 .BI "mount -t lustre [-o " options "] " "<mgsname>:/<fsname> <client_mountpoint>"
16 .BI "mount -t lustre_tgt [-o " options "] " "<block_device> <server_mountpoint>"
21 are used to start a Lustre client or target service, respectively. They
22 should not be called directly; rather they are helper programs invoked through
28 filesystem type is used. Lustre clients and targets are stopped by using the
33 There are two forms for the device option passed to the
35 command, depending on whether a client or a target service is started:
37 .IR <mgsname> :/ <fsname>[/<subdir>]
38 mounts the Lustre client filesystem named
40 (optionally starting at subdirectory
42 within the filesystem, if specified) on the client at the directory
43 .IR client_mountpoint ,
44 by contacting the Management Service at
48 is defined below. A client filesystem will typically be listed in
50 for automatic mount at boot time, is usable like any local filesystem, and
51 provides a full POSIX-compilant interface.
54 specifies the phyisical disk or ZFS dataset uses to start the target service
59 may be specified using
61 to find the first block device with that label (e.g.
65 option. Care should be taken if there is a device-level backup of
66 the target filesystem, which would have a duplicate label and UUID if it is
69 or similar. The mounted target service filesystem at
73 operations to show the current space usage, and appears in
75 to show the device is in use.
78 .BI <mgsname>:= <mgsnode>[:<mgsnode>]
81 may be a colon-separated list of
83 names where the MGS service may run. Multiple
85 values can be specified if the MGS service is configures for HA failover
86 and may be running on any one of the nodes.
88 .BI <mgsnode>:= <mgsnid>[,<mgsnid>]
91 may specify a comma-separated list of NIDs, if there are different
92 LNet interfaces for the same
96 Specifies the encryption flavour for the initial network RPC connection to
97 the MGS node. Non-security flavors are:
102 which respectively disable, or have no encryption or integrity features for
103 testing purposes. Kerberos flavors are:
109 Shared-secret key flavors are:
117 for more details. The security flavour for client-to-server connections is
118 specified in the filesystem configuration that the client fetches from the MGS.
120 .BI skpath= file|directory
121 Path to a file or directory with the keyfile(s) to load for this mount command.
122 Keys are inserted into the KEY_SPEC_SESSION_KEYRING keyring with a description
123 containing "lustre:" and a suffix which depends on whether the context of the
124 mount command is for an MGS, MDT/OST, or client.
125 This option is only available when built with --enable-gss.
128 Start a client or MDT with a (colon-separated) list of known inactive OSTs.
130 In addition to the standard options listed in
132 Lustre understands the following
137 Enable full distributed
139 support, coherent across all client nodes also using this mount option. This
140 is useful if applications need coherent userspace file locking across multiple
141 client nodes, has no measurable performance impact in modern (2.x) versions of
142 Lustre, is therefore the recommended mode, and is enabled by default in Lustre
148 support, using only client-local file locking. This is faster than mounting
151 option, and can be used for applications that depend on functioning
153 but run only on a single node.
158 support entirely. Applications calling
162 error. It is possible to mount clients
163 with different options, and only those mounted with
165 will be coherent amongst each other.
174 to return even if some OST or MDT is unresponsive or has been temporarily
175 or permanently disabled in the configuration. This avoids blocking until
176 all of the targets are available. This is the default since Lustre 2.9.0.
181 block until all OSTs and MDTs are available and have returned space usage.
184 Enable get/set of extended attributes by regular users. See the
189 Disable use of extended attributes by regular users. Root and system processes can still use extended attributes.
192 Force a client to keep pinging even if servers have enabled suppress_pings.
195 Enable mount/remount/umount console messages.
198 Disable mount/remount/umount console messages.
201 Enable FID to path translation by regular users via
205 Disable FID to path translation by regular users. Root and process with
206 CAP_DAC_READ_SEARCH can still perform FID to path translation.
209 Limit connections from the client to be on the network NID specified by 'net'.
210 \'net\' designates a single network NID, like 'o2ib2' or 'tcp1'.
211 This option can be useful in case of several Lustre client mount
212 points on the same node, with each mount point using a different
213 network. It is also interesting when running Lustre clients from
214 containers, by restricting each container to a specific network.
217 Warning! 'network' option is incompatible with LNet Dynamic Peer Discovery.
218 If you want to restrict client NID, please make sure LNet Dynamic Peer Discovery
222 .BI test_dummy_encryption
223 Enable test dummy encryption mode.
227 Disable Lustre client-side encryption. By default, Lustre client-side encryption
228 is enabled, letting users define encryption policies on a per-directory basis.
229 fscrypt userspace tool can be used for that purpose, see
230 https://github.com/google/fscrypt
232 In addition to the standard mount options and backing disk type
233 (e.g. ldiskfs) options listed in
235 Lustre understands the following
240 Enable POSIX Access Control List support for all clients. See the
245 Do not allow objects to be created on the OST after mounting. This
246 allows a (potentially newly-formatted) OST to be started without
247 clients starting to use it immediately. To allow the MDS to create
248 objects on the OST mount without this option, or run
249 .BI "lctl set_param obdfilter." fsname -OST nnnn .no_precreate=0
250 on the OSS to re-enable object creation.
252 this does not start the OST read-only, it only disables the
253 creation of new objects on the OST. This means existing objects
254 can be read, modified, or deleted, but if the OST is new then it
255 will not have any such objects.
258 Only start the MGC (and MGS, if co-located) for a target service,
259 and not the actual MDS or OSS service.
262 Start an MDT with a co-located MGS without starting the MGS service.
265 Do not trigger OI scrub automatically when an inconsistency is detected.
266 It is still possible to start explicitly using the
267 .BR "lctl lfsck_start" .
270 Do not resume the former paused/crashed LFSCK automatically when mounting.
273 Abort client recovery and start the target service immediately.
275 .BI md_stripe_cache_size
276 Sets the stripe cache size for server side disk with a striped raid
280 Automatically Sets the block device parameter of 'max_sectors_kb' for the
281 MDT or OST target. When max_sectors_kb isn't specified, that parameter for
282 block device will be set to same as it's own 'max_hw_sectors_kb' (up to a
283 maximum of 16M), this is default behavior suited for most users. When
284 max_sectors_kb is specified as zero, the old parameter value will be kept.
285 When max_sectors_kb is specified as a positive number, the parameter will
286 be set to this number arbitrarily.
288 .BI recovery_time_soft= timeout
289 Allow 'timeout' seconds for clients to reconnect for recovery after a server
290 crash. This timeout will be incrementally extended if it is about to expire
291 and the server is still handling new connections from recoverable clients.
292 The default soft recovery timeout is set to 300 seconds (5 minutes).
294 .BI recovery_time_hard= timeout
295 The server will be allowed to incrementally extend its timeout up to a hard
296 maximum of 'timeout' seconds. The default hard recovery timeout is set to
297 900 seconds (15 minutes).
300 .B mount -t lustre cfs21@tcp0:/testfs /mnt/myfilesystem
301 Start a client for the Lustre filesystem
304 .BR /mnt/myfilesystem .
305 The Management Service is running on a node reachable via NID
308 .B mount -t lustre cfs21@tcp0:/testfs/dir /mnt/myfilesystem
309 Like above example, but mount subdirectory
313 .B mount -t lustre mgs1@tcp0,mgs1ib@o2ib0:mgs2@tcp0,mgs2ib@o2ib0:/testfs /mnt/fs
314 Like above example, but the Management Service is running on one of the service
318 .B mgs2, which are two different hosts separated by a colon and
319 served as a failover pair. Lustre tries the first one, and if that fails, it
320 tries the second one. On each service node, the comma-separated NIDs refer to
321 different interfaces on the same host, and the Lustre client chooses the best
322 one for communication based on which network interfaces are available locally.
324 .B mount -t lustre_tgt /dev/sda1 /mnt/test/mdt
325 Start the Lustre metadata target service from
330 .B mount -t lustre_tgt -L testfs-MDT0000 -o abort_recov /mnt/test/mdt
333 service (by using the disk label), but aborts the Lustre client recovery
334 process for the case that clients are known to be unavailable.
336 Not very many mount options can be changed with
347 .BR tunefs.lustre (8),