<para>This chapter describes how to configure quotas and includes the following sections:</para>
<itemizedlist>
<listitem>
- <para><xref linkend="dbdoclet.50438217_54945"/></para>
+ <para><xref linkend="quota_configuring"/></para>
</listitem>
<listitem>
- <para><xref linkend="dbdoclet.50438217_31982"/></para>
+ <para><xref linkend="enabling_disk_quotas"/></para>
</listitem>
<listitem>
- <para><xref linkend="dbdoclet.50438217_49939"/></para>
+ <para><xref linkend="quota_administration"/></para>
</listitem>
<listitem>
- <para><xref linkend="dbdoclet.50438217_15106"/></para>
+ <para><xref linkend="quota_allocation"/></para>
</listitem>
<listitem>
- <para><xref linkend="dbdoclet.50438217_27895"/></para>
+ <para><xref linkend="quota_interoperability"/></para>
</listitem>
<listitem>
- <para><xref linkend="dbdoclet.50438217_20772"/></para>
+ <para><xref linkend="granted_cache_and_quota_limits"/></para>
+ </listitem>
+ <listitem>
+ <para><xref linkend="lustre_quota_statistics"/></para>
</listitem>
</itemizedlist>
- <section xml:id="dbdoclet.50438217_54945">
+ <section xml:id="quota_configuring">
<title>
- <indexterm><primary>Quotas</primary><secondary>configuring</secondary></indexterm>
- Working with Quotas</title>
+ <indexterm><primary>Quotas</primary><secondary>configuring</secondary></indexterm>Working with Quotas</title>
<para>Quotas allow a system administrator to limit the amount of disk space a user or group can use. Quotas are set by root, and can be specified for individual users and/or groups. Before a file is written to a partition where quotas are set, the quota of the creator's group is checked. If a quota exists, then the file size counts towards the group's quota. If no quota exists, then the owner's user quota is checked before the file is written. Similarly, inode usage for specific functions can be controlled if a user over-uses the allocated space.</para>
<para>Lustre quota enforcement differs from standard Linux quota enforcement in several ways:</para>
<itemizedlist>
<listitem>
- <para>Quotas are administered via the <literal>lfs</literal> command (post-mount).</para>
+ <para>Quotas are administered via the <literal>lfs</literal> and <literal>lctl</literal> commands (post-mount).</para>
</listitem>
<listitem>
<para>Quotas are distributed (as Lustre is a distributed file system), which has several ramifications.</para>
<para>Quotas are allocated and consumed in a quantized fashion.</para>
</listitem>
<listitem>
- <para>Client does not set the <literal>usrquota</literal> or <literal>grpquota</literal> options to mount. When quota is enabled, it is enabled for all clients of the file system; started automatically using <literal>quota_type</literal> or started manually with <literal>lfs quotaon</literal>.</para>
+ <para>Client does not set the <literal>usrquota</literal> or <literal>grpquota</literal> options to mount. As of Lustre 2.4, space accounting is always enabled by default and quota enforcement can be enabled/disabled on a per-filesystem basis with <literal>lctl conf_param</literal>. It is worth noting that both <literal>lfs quotaon</literal> and <literal>quota_type</literal> are deprecated as of Lustre 2.4.0.</para>
</listitem>
</itemizedlist>
<caution>
<para><literal>lfs quota -u root</literal> (usage includes internal Lustre data that is dynamic in size and does not accurately reflect mount point visible block and inode usage).</para>
</caution>
</section>
- <section xml:id="dbdoclet.50438217_31982">
+ <section xml:id="enabling_disk_quotas">
<title><indexterm><primary>Quotas</primary><secondary>enabling disk</secondary></indexterm>Enabling Disk Quotas</title>
- <para>Use this procedure to enable (configure) disk quotas in Lustre.</para>
- <orderedlist>
+ <para>Prior to Lustre 2.4.0, enabling quota involved a full filesystem scan via <literal>lfs quotacheck</literal>. All filesystems formatted with Lustre 2.4.0 or newer no longer require quotacheck to be run since up-to-date accounting information are now always maintained by the OSD layer, regardless of the quota enforcement status. Although quota enforcement is managed by Lustre itself, each OSD implementation relies on the backend filesystem to maintain per-user/group block and inode usage:</para>
+ <itemizedlist>
+ <listitem>
+ <para>For ldiskfs backend, mkfs.lustre now creates empty quota files and enables the QUOTA feature flag in the superblock which turns quota accounting on at mount time automatically. e2fsck was also modified to fix the quota files when the QUOTA feature flag is present.</para>
+ </listitem>
+ <listitem>
+ <para>For ZFS backend, accounting ZAPs are created and maintained by the ZFS filesystem itself. While ZFS tracks per-user and group block usage, it does not handle inode accounting. The ZFS OSD implements its own support for inode tracking. Two options are available:</para>
+ <orderedlist>
+ <listitem>
+ <para> The ZFS OSD can estimate the number of inodes in-use based on the number of blocks used by a given user or group. This mode can be enabled by running the following command on the server running the target: <literal>lctl set_param osd-zfs.${FSNAME}-${TARGETNAME}.quota_iused_estimate=1</literal>.</para>
+ </listitem>
+ <listitem>
+ <para> Similarly to block accounting, dedicated ZAPs are also created the ZFS OSD to maintain per-user and group inode usage. This is the default mode which corresponds to <literal>quota_iused_estimate</literal> set to 0.</para>
+ </listitem>
+ </orderedlist>
+ </listitem>
+ </itemizedlist>
+ <para>As a result, <literal>lfs quotacheck</literal> is now deprecated and not required any more when running Lustre 2.4 on the servers.</para>
+ <para>Lustre filesystems formatted with a Lustre version prior to 2.4.0 can be still safely upgraded to 2.4.0, but won't have functional space usage report until <literal>tunefs.lustre --quota</literal> is run against all targets. This command sets the QUOTA feature flag in the superblock and runs e2fsck (as a result, the target must be offline) to build the per-UID/GID disk usage database. </para>
+
+ <caution>
+ <para>Lustre 2.4 and beyond requires a version of e2fsprogs that supports quota (i.e. newer or equal to 1.42.3.wc1) to be installed on the server nodes using ldiskfs backend (e2fsprogs isn't needed with ZFS backend). In general, we recommend to use the latest e2fsprogs version available on http://downloads.whamcloud.com/public/e2fsprogs.</para>
+ <para>The ldiskfs OSD relies on the standard Linux quota to maintain accounting information on disk. As a consequence, the Linux kernel running on the Lustre servers using ldiskfs backend must have <literal>CONFIG_QUOTA</literal>, <literal>CONFIG_QUOTACTL</literal> and <literal>CONFIG_QFMT_V2</literal> enabled.</para>
+ </caution>
+
+ <para>As of Lustre 2.4.0, quota enforcement is thus turned on/off independently of space accounting which is always enabled. <literal>lfs quota<on|off></literal> as well as the per-target <literal>quota_type</literal> parameter are deprecated in favor of a single per-filesystem quota parameter controlling inode/block quota enforcement. Like all permanent parameters, this quota parameter can be set via <literal>lctl conf_param</literal> on the MGS via the following syntax:</para>
+ <screen>lctl conf_param ${FSNAME}.quota.<ost|mdt>=<u|g|ug|none></screen>
+ <itemizedlist>
<listitem>
- <para>If you have re-complied your Linux kernel, be sure that <literal>CONFIG_QUOTA</literal> and <literal>CONFIG_QUOTACTL</literal> are enabled. Also, verify that <literal>CONFIG_QFMT_V1</literal> and/or <literal>CONFIG_QFMT_V2</literal> are enabled.</para>
- <para>Quota is enabled in all Linux 2.6 kernels supplied for Lustre.</para>
+ <para><literal>ost</literal> -- to configure block quota managed by OSTs</para>
</listitem>
<listitem>
- <para>Start the server.</para>
+ <para><literal>mdt</literal> -- to configure inode quota managed by MDTs</para>
</listitem>
<listitem>
- <para>
- Mount the Lustre file system on the client and verify that the <literal>lquota</literal> module has loaded properly by using the <literal>lsmod</literal> command.
-
- </para>
- <screen>$ lsmod
-[root@oss161 ~]# lsmod
-Module Size Used by
-obdfilter 220532 1
-fsfilt_ldiskfs 52228 1
-ost 96712 1
-mgc 60384 1
-ldiskfs 186896 2 fsfilt_ldiskfs
-lustre 401744 0
-lov 289064 1 lustre
-lquota 107048 4 obdfilter
-mdc 95016 1 lustre
-ksocklnd 111812 1</screen>
+ <para><literal>u</literal> -- to enable quota enforcement for users only</para>
</listitem>
- </orderedlist>
- <para>The Lustre mount command no longer recognizes the <literal>usrquota</literal> and <literal>grpquota</literal> options. If they were previously specified, remove them from <literal>/etc/fstab</literal>.</para>
- <para>When quota is enabled, it is enabled for all file system clients (started automatically using <literal>quota_type</literal> or manually with <literal>lfs quotaon</literal>).</para>
- <para>To enable quotas automatically when the file system is started, you must set the <literal>mdd.quota_type</literal> and <literal>ost.quota_type</literal> parameters, respectively, on the MDT and OSTs. The parameters can be set to the string <literal>u</literal> (user), <literal>g</literal> (group) or <literal>ug</literal> for both users and groups.</para>
- <para>You can enable quotas at <literal>mkfs</literal> time (<literal>mkfs.lustre --param mdd.quota_type=ug</literal>) or with <literal>tunefs.lustre</literal>. As an example:</para>
- <screen>tunefs.lustre --param ost.quota_type=ug $ost_dev</screen>
+ <listitem>
+ <para><literal>g</literal> -- to enable quota enforcement for groups only</para>
+ </listitem>
+ <listitem>
+ <para><literal>ug</literal> -- to enable quota enforcement for both users and groups</para>
+ </listitem>
+ <listitem>
+ <para><literal>none</literal> -- to disable quota enforcement for both users and groups</para>
+ </listitem>
+ </itemizedlist>
+
+ <para>Examples:</para>
+ <para>To turn on user and group quotas for block only on filesystem testfs1, run:</para>
+ <screen>$ lctl conf_param testfs1.quota.ost=ug</screen>
+ <para>To turn on group quotas for inodes on filesystem testfs2, run:</para>
+ <screen>$ lctl conf_param testfs2.quota.mdt=g</screen>
+ <para>To turn off user and group quotas for both inode and block on filesystem testfs3, run:</para>
+ <screen>$ lctl conf_param testfs3.quota.ost=none</screen>
+ <screen>$ lctl conf_param testfs3.quota.mdt=none</screen>
+
+ <para>Once the quota parameter set on the MGS, all targets which are part of the filesystem will be notified of the new quota settings and enable/disable quota enforcement as needed. The per-target enforcement status can still be verified by running the following command on the Lustre servers:</para>
+ <screen>$ lctl get_param osd-*.*.quota_slave.info
+osd-zfs.testfs-MDT0000.quota_slave.info=
+target name: testfs-MDT0000
+pool ID: 0
+type: md
+quota enabled: ug
+conn to master: setup
+user uptodate: glb[1],slv[1],reint[0]
+group uptodate: glb[1],slv[1],reint[0]</screen>
<caution>
- <para>If you are using <literal>mkfs.lustre --param mdd.quota_type=ug</literal> or <literal>tunefs.lustre --param ost.quota_type=ug</literal>, be sure to run the command on all OSTs and the MDT. Otherwise, abnormal results may occur.</para>
+ <para>Lustre 2.4 comes with a new quota protocol and a new on-disk format, be sure to check the Interoperability section below (see <xref linkend="quota_interoperability"/>.) when migrating to 2.4</para>
</caution>
- <section remap="h4">
- <title><indexterm><primary>Quotas</primary><secondary>administrating</secondary></indexterm>Administrative and Operational Quotas</title>
- <para>Lustre has two kinds of quota files:</para>
- <itemizedlist>
- <listitem>
- <para>Administrative quotas (for the MDT), which contain limits for users/groups for the entire cluster.</para>
- </listitem>
- <listitem>
- <para>Operational quotas (for the MDT and OSTs), which contain quota information dedicated to a cluster node.</para>
- </listitem>
- </itemizedlist>
- <para>Lustre 1.6.5 introduced the v2 file format for administrative quota files, with continued support for the old file format (v1). The mdd.quota_type parameter also handles '1' and '2' options, to specify the Lustre quota versions that will be used. For example:</para>
- <screen>--param mdd.quota_type=ug1
---param mdd.quota_type=u2</screen>
- <para>Lustre 1.6.6 introduced the v2 file format for operational quotas, with continued support for the old file format (v1). The ost.quota_type parameter handles '1' and '2' options, to specify the Lustre quota versions that will be used. For example:</para>
- <screen>--param ost.quota_type=ug2
---param ost.quota_type=u1</screen>
- <para>For more information about the v1 and v2 formats, see <xref linkend="dbdoclet.50438217_66360"/>.</para>
- </section>
- </section>
- <section xml:id="dbdoclet.50438217_49939">
- <title><indexterm><primary>Quotas</primary><secondary>creating</secondary></indexterm>Creating Quota Files and Quota Administration</title>
- <para>Once each quota-enabled file system is remounted, it is capable of working with disk quotas. However, the file system is not yet ready to support quotas. If <literal>umount</literal> has been done regularly, run the <literal>lfs</literal> command with the <literal>quotaon</literal> option. If <literal>umount</literal> has not been done, perform these steps:</para>
- <orderedlist>
+
+ <para>For Lustre versions older than 2.4, <literal>lfs quotacheck</literal> must be first run from a client node to create quota files on the Lustre targets (i.e. the MDT and OSTs). <literal>lfs quotacheck</literal> requires the filesystem to be quiescent (i.e. no modifying operations like write, truncate, create or delete should run concurrently). Failure to follow this caution may result in inaccurate user/group disk usage. Operations that do not change Lustre files (such as read or mount) are okay to run. <literal>lfs quotacheck</literal> performs a scan on all the Lustre targets to calculates the block/inode usage for each user/group. If the Lustre filesystem has many files, <literal>quotacheck</literal> may take a long time to complete. Several options can be passed to <literal>lfs quotacheck</literal>:</para>
+ <screen># lfs quotacheck -ug /mnt/testfs</screen>
+ <itemizedlist>
<listitem>
- <para>Take Lustre ''offline''.</para>
- <para>That is, verify that no write operations (append, write, truncate, create or delete) are being performed (preparing to run <literal>lfs quotacheck</literal>). Operations that do not change Lustre files (such as read or mount) are okay to run.</para>
- <caution>
- <para>When <literal>lfs quotacheck</literal> is run, Lustre must NOT be performing any write operations. Failure to follow this caution may cause the statistic information of quota to be inaccurate. For example, the number of blocks used by OSTs for users or groups will be inaccurate, which can cause unexpected quota problems.</para>
- </caution>
+ <para><literal>u</literal> -- checks the user disk quota information</para>
</listitem>
<listitem>
- <para> Run the <emphasis role="bold">
- <literal>lfs</literal>
- </emphasis> command with the <emphasis role="bold">
- <literal>quotacheck</literal>
- </emphasis> option:</para>
- <screen># lfs quotacheck -ug /mnt/lustre</screen>
- <para>By default, quota is turned on after <literal>quotacheck</literal> completes. Available options are:</para>
- <itemizedlist>
- <listitem>
- <para><literal>u</literal> -- checks the user disk quota information</para>
- </listitem>
- <listitem>
- <para><literal>g</literal> -- checks the group disk quota information</para>
- </listitem>
- </itemizedlist>
+ <para><literal>g</literal> -- checks the group disk quota information</para>
</listitem>
- </orderedlist>
- <para>The lfsquotacheck command checks all objects on all OSTs and the MDS to sum up for every UID/GID. It reads all Lustre metadata and re-computes the number of blocks/inodes that each UID/GID has used. If there are many files in Lustre, it may take a long time to complete.</para>
- <note>
- <para>User and group quotas are separate. If either quota limit is reached, a process with the corresponding UID/GID cannot allocate more space on the file system.</para>
- </note>
- <note>
- <para>When <literal>lfs quotacheck</literal> runs, it creates a quota file -- a sparse file with a size proportional to the highest UID in use and UID/GID distribution. As a general rule, if the highest UID in use is large, then the sparse file will be large, which may affect functions such as creating a snapshot.</para>
- </note>
- <note>
- <para>Lustre 1.6.5 and later use journaled quota files, so it is not necessary to run <literal>quotacheck</literal> after an unclean shutdown.</para>
- <para>In certain failure situations (e.g., when a broken Lustre installation or build is used), re-run <literal>quotacheck</literal> after checking the server kernel logs and fixing the root problem.</para>
- </note>
- <para>The <literal>lfs</literal> command includes several command options to work with quotas:</para>
+ </itemizedlist>
+ <para>By default, quota is turned on after <literal>quotacheck</literal> completes. However, this setting isn't persistent and quota will have to be enabled again (via <literal>lfs quotaon</literal>) if one of the Lustre target is restarted. <literal>lfs quotaoff</literal> should be used to turn off quota.</para>
+ <para>To enable quota permanently with a Lustre version older than 2.4, the <literal>quota_type</literal> parameter must be used. This requires to set <literal>mdd.quota_type</literal> and <literal>ost.quota_type</literal>, respectively, on the MDT and OSTs. <literal>quota_type</literal> can be set to the string <literal>u</literal> (user), <literal>g</literal> (group) or <literal>ug</literal> for both users and groups. This parameter can be specified at <literal>mkfs</literal> time (<literal>mkfs.lustre --param mdd.quota_type=ug</literal>) or with <literal>tunefs.lustre</literal>. As an example:</para>
+ <screen>tunefs.lustre --param ost.quota_type=ug $ost_dev</screen>
+ <para>When using <literal>mkfs.lustre --param mdd.quota_type=ug</literal> or <literal>tunefs.lustre --param ost.quota_type=ug</literal>, be sure to run the command on all OSTs and the MDT. Otherwise, abnormal results may occur.</para>
+
+ </section>
+ <section xml:id="quota_administration">
+ <title><indexterm><primary>Quotas</primary><secondary>creating</secondary></indexterm>Quota Administration</title>
+ <para>Once the filesystem is up and running, quota limits on blocks and files can be set for both user and group. This is controlled via three quota parameters:</para>
+ <para><emphasis role="bold">Grace period</emphasis> -- The period of time (in seconds) within which users are allowed to exceed their soft limit. There are four types of grace periods:</para>
<itemizedlist>
<listitem>
- <para><varname>quotaon</varname> -- enables disk quotas on the specified file system. The file system quota files must be present in the root directory of the file system.</para>
+ <para> user block soft limit</para>
</listitem>
<listitem>
- <para><varname>quotaoff</varname> -- disables disk quotas on the specified file system.</para>
+ <para> user inode soft limit</para>
</listitem>
<listitem>
- <para><varname>quota</varname> -- displays general quota information (disk usage and limits)</para>
+ <para> group block soft limit</para>
+ </listitem>
+ <listitem>
+ <para> group inode soft limit</para>
+ </listitem>
+ </itemizedlist>
+ <para>The grace period applies to all users. The user block soft limit is for all users who are using a blocks quota.</para>
+ <para><emphasis role="bold">Soft limit</emphasis> -- The grace timer is started once the soft limit is exceeded. At this point, the user/group can still allocate block/inode. When the grace time expires and if the user is still above the soft limit, the soft limit becomes a hard limit and the user/group can't allocate any new block/inode any more. The user/group should then delete files to be under the soft limit. The soft limit MUST be smaller than the hard limit. If the soft limit is not needed, it should be set to zero (0).</para>
+ <para><emphasis role="bold">Hard limit</emphasis> -- Block or inode allocation will fail with <literal>EDQUOT</literal> (i.e. quota exceeded) when the hard limit is reached. The hard limit is the absolute limit. When a grace period is set, one can exceed the soft limit within the grace period if under the hard limit.</para>
+ <para>Due to the distributed nature of Lustre and the need to mainain performance under load, those quota parameters may not be 100% accurate. The quota settings can be manipulated via the <literal>lfs</literal> command which includes several options to work with quotas:</para>
+ <itemizedlist>
+ <listitem>
+ <para><varname>quota</varname> -- displays general quota information (disk usage and limits)</para>
</listitem>
<listitem>
<para><varname>setquota</varname> -- specifies quota limits and tunes the grace period. By default, the grace period is one week.</para>
</listitem>
</itemizedlist>
<para> Usage:</para>
- <screen>lfs quotaon [-ugf] <filesystem>
-lfs quotaoff [-ug] <filesystem>
-lfs quota [-q] [-v] [-o obd_uuid] [-u|-g <uname>|uid|gname|gid>] <filesystem>
+ <screen>lfs quota [-q] [-v] [-o obd_uuid] [-u|-g <uname>|uid|gname|gid>] <filesystem>
lfs quota -t <-u|-g> <filesystem>
-lfs setquota <-u|--user|-g|--group> <username|groupname> [-b <block-softlimit>] [\
--B <block-hardlimit>] [-i <inode-softlimit>] [-I <inode-hardlimit>] <filesystem></screen>
- <para>Examples:</para>
- <para>In all of the examples below, the file system is <literal>/mnt</literal> lustre.</para>
- <para>To turn on user and group quotas, run:</para>
- <screen>$ lfs quotaon -ug /mnt/lustre</screen>
- <para>To turn off user and group quotas, run:</para>
- <screen>$ lfs quotaoff -ug /mnt/lustre</screen>
+lfs setquota <-u|--user|-g|--group> <username|groupname> [-b <block-softlimit>] \
+ [-B <block-hardlimit>] [-i <inode-softlimit>] \
+ [-I <inode-hardlimit>] <filesystem></screen>
<para>To display general quota information (disk usage and limits) for the user running the command and his primary group, run:</para>
- <screen>$ lfs quota /mnt/lustre </screen>
+ <screen>$ lfs quota /mnt/testfs </screen>
<para>To display general quota information for a specific user ("<literal>bob</literal>" in this example), run:</para>
- <screen>$ lfs quota -u bob /mnt/lustre</screen>
+ <screen>$ lfs quota -u bob /mnt/testfs</screen>
<para>To display general quota information for a specific user ("<literal>bob</literal>" in this example) and detailed quota statistics for each MDT and OST, run:</para>
- <screen>$ lfs quota -u bob -v /mnt/lustre</screen>
+ <screen>$ lfs quota -u bob -v /mnt/testfs</screen>
<para>To display general quota information for a specific group ("<literal>eng</literal>" in this example), run:</para>
- <screen>$ lfs quota -g eng /mnt/lustre</screen>
+ <screen>$ lfs quota -g eng /mnt/testfs</screen>
<para>To display block and inode grace times for user quotas, run:</para>
- <screen>$ lfs quota -t -u /mnt/lustre</screen>
- <para>To set user and group quotas for a specific user ("bob" in this example), run:</para>
- <screen>$ lfs setquota -u bob 307200 309200 10000 11000 /mnt/lustre</screen>
+ <screen>$ lfs quota -t -u /mnt/testfs</screen>
+ <para>To set user or group quotas for a specific ID ("bob" in this example), run:</para>
+ <screen>$ lfs setquota -u bob -b 307200 -B 309200 -i 10000 -I 11000 /mnt/testfs</screen>
<para>In this example, the quota for user "bob" is set to 300 MB (309200*1024) and the hard limit is 11,000 files. Therefore, the inode hard limit should be 11000.</para>
- <note>
- <para>For the Lustre command <literal>$lfssetquota/quota ...</literal> the qunit for block is KB (1024) and the qunit for inode is 1.</para>
- </note>
- <para>The quota command displays the quota allocated and consumed for each Lustre device. Using the previous <literal>setquota</literal> example, running this <literal>lfs</literal> quota command:</para>
- <screen>$ lfs quota -u bob -v /mnt/lustre </screen>
+ <para>The quota command displays the quota allocated and consumed by each Lustre target. Using the previous <literal>setquota</literal> example, running this <literal>lfs</literal> quota command:</para>
+ <screen>$ lfs quota -u bob -v /mnt/testfs </screen>
<para>displays this command output:</para>
<screen>Disk quotas for user bob (uid 6000):
-Filesystem kbytes quota limit grace \
- files quota limit grace
-/mnt/lustre 0 30720 30920 \
-- 0 10000 11000 -
-lustre-MDT0000_UUID 0 - 16384 \
-- 0 - 2560 -
-lustre-OST0000_UUID 0 - 16384 \
-- 0 - 0 -
-lustre-OST0001_UUID 0 - 16384 \
-- 0 - 0 -</screen>
+Filesystem kbytes quota limit grace files quota limit grace
+/mnt/testfs 0 30720 30920 - 0 10000 11000 -
+testfs-MDT0000_UUID 0 - 16384 - 0 - 2560 -
+testfs-OST0000_UUID 0 - 16384 - 0 - 0 -
+testfs-OST0001_UUID 0 - 16384 - 0 - 0 -</screen>
+ <para>Global quota limits are stored in dedicated index files (there is one such index per quota type) on the quota master target (aka QMT). The QMT runs on MDT0000 and exports the global indexes via /proc. The global indexes can thus be dumped via the following command: <screen># lctl get_param qmt.testfs-QMT0000.*.glb-*</screen> The format of global indexes depends on the OSD type. The ldiskfs OSD uses an IAM files while the ZFS OSD creates dedicated ZAPs.</para>
+ <para>Each slave also stores a copy of this global index locally. When the global index is modified on the master, a glimpse callback is issued on the global quota lock to notify all slaves that the global index has been modified. This glimpse callback includes information about the identifier subject to the change. If the global index on the QMT is modified while a slave is disconnected, the index version is used to determine whether the slave copy of the global index isn't uptodate any more. If so, the slave fetches the whole index again and updates the local copy. The slave copy of the global index is also exported via /proc and can be accessed via the following command: <screen>lctl get_param osd-*.*.quota_slave.limit*</screen></para>
+ <note>
+ <para>Prior to 2.4, global quota limits used to be stored in administrative quota files using the on-disk format of the linux quota file. When upgrading MDT0000 to 2.4, those administrative quota files are converted into IAM indexes automatically, conserving existing quota limits previously set by the administrator.</para>
+ </note>
</section>
- <section xml:id="dbdoclet.50438217_15106">
+ <section xml:id="quota_allocation">
<title><indexterm><primary>Quotas</primary><secondary>allocating</secondary></indexterm>Quota Allocation</title>
- <para>In Lustre, quota must be properly allocated or users may experience unnecessary failures. The file system block quota is divided up among the OSTs within the file system. Each OST requests an allocation which is increased up to the quota limit. The quota allocation is then <emphasis role="italic">quantized</emphasis> to reduce the number of quota-related request traffic. By default, Lustre supports both user and group quotas to limit disk usage and file counts.</para>
- <para>The quota system in Lustre is completely compatible with the quota systems used on other file systems. The Lustre quota system distributes quotas from the quota master. Generally, the MDS is the quota master for both inodes and blocks. All OSTs and the MDS are quota slaves to the OSS nodes. To reduce quota requests and get reasonably accurate quota distribution, the transfer quota unit (qunit) between quota master and quota slaves is changed dynamically by the lquota module. The default minimum value of qunit is 1 MB for blocks and 2 for inodes. The proc entries to set these values are: <literal>/proc/fs/lustre/mds/lustre-MDT*/quota_least_bunit</literal> and <literal>/proc/fs/lustre/mds/lustre-MDT*/quota_least_iunit</literal>. The default maximum value of <literal>qunit</literal> is 128 MB for blocks and 5120 for inodes. The proc entries to set these values are <literal>quota_bunit_sz</literal> and <literal>quota_iunit_sz</literal> in the MDT and OSTs.</para>
- <note>
- <para>In general, the <literal>quota_bunit_sz</literal> value should be larger than 1 MB. For testing purposes, it can be set to 4 KB, if necessary.</para>
- </note>
- <para>The file system block quota is divided up among the OSTs and the MDS within the file system. Only the MDS uses the file system inode quota.</para>
- <para>This means that the minimum quota for block is 1 MB* (the number of OSTs + the number of MDSs), which is 1 MB* (number of OSTs + 1). If you attempt to assign a smaller quota, users maybe not be able to create files. As noted, the default minimum quota for inodes is 2. The default is established at file system creation time, but can be tuned via <literal>/proc</literal> values (described below). The inode quota is also allocated in a quantized manner on the MDS.</para>
+ <para>In Lustre, quota must be properly allocated or users may experience unnecessary failures. The file system block quota is divided up among the OSTs within the file system. Each OST requests an allocation which is increased up to the quota limit. The quota allocation is then <emphasis role="italic">quantized</emphasis> to reduce the number of quota-related request traffic.</para>
+ <para>The Lustre quota system distributes quotas from the Quota Master Target (aka QMT). Only one QMT instance is supported for now and only runs on the same node as MDT0000. All OSTs and MDTs set up a Quota Slave Device (aka QSD) which connects to the QMT to allocate/release quota space. The QSD is setup directly from the OSD layer.</para>
+ <para>To reduce quota requests, quota space is initially allocated to QSDs in very large chunks. How much unused quota space can be hold by a target is controlled by the qunit size. When quota space for a given ID is close to exhaustion on the QMT, the qunit size is reduced and QSDs are notified of the new qunit size value via a glimpse callback. Slaves are then responsible for releasing quota space above the new qunit value. The qunit size isn't shrunk indefinitely and there is a minimal value of 1MB for blocks and 1,000 for inodes. This means that the quota space rebalancing process will stop when this mininum value is reached. As a result, quota exceeded can be returned while many slaves still have 1MB or 1,000 inodes of spare quota space.</para>
<para>If we look at the <literal>setquota</literal> example again, running this <literal>lfs quota</literal> command:</para>
- <screen># lfs quota -u bob -v /mnt/lustre
+ <screen># lfs quota -u bob -v /mnt/testfs
</screen>
<para>displays this command output:</para>
<screen>Disk quotas for user bob (uid 500):
-Filesystem kbytes quota limit grace \
- files quota limit grace
-/mnt/lustre 30720* 30720 30920 \
-6d23h56m44s 10101* 10000 11000 6d23h59m50s
-lustre-MDT0000_UUID 0 - 1024 \
-- 10101 - 10240
-lustre-OST0000_UUID 0 - 1024 \
-- - - -
-lustre-OST0001_UUID 30720* - 28872 \
-- - - -</screen>
- <para>The total quota limit of 30,920 is allotted to user bob, which is further distributed to two OSTs and one MDS.</para>
- <note>
- <para>Values appended with '<literal>*</literal>' show the limit that has been over-used (exceeding the quota), and receives this message Disk quota exceeded. For example:</para>
- <para><screen>$ cp: writing `/mnt/lustre/var/cache/fontconfig/ beeeeb3dfe132a8a0633a017c99ce0-x86.cache': Disk quota exceeded.</screen></para>
- </note>
- <para>The requested quota of 300 MB is divided across the OSTs.</para>
+Filesystem kbytes quota limit grace files quota limit grace
+/mnt/testfs 30720* 30720 30920 6d23h56m44s 10101* 10000 11000 6d23h59m50s
+testfs-MDT0000_UUID 0 - 0 - 10101 - 10240
+testfs-OST0000_UUID 0 - 1024 - - - -
+testfs-OST0001_UUID 30720* - 29896 - - - -</screen>
+ <para>The total quota limit of 30,920 is allocated to user bob, which is further distributed to two OSTs.</para>
+ <para>Values appended with '<literal>*</literal>' show that the quota limit has been exceeded, causing the following error when trying to write or create a file:</para>
+ <para><screen>$ cp: writing `/mnt/testfs/foo`: Disk quota exceeded.</screen></para>
<note>
- <para>It is very important to note that the block quota is consumed per OST and the MDS per block and inode (there is only one MDS for inodes). Therefore, when the quota is consumed on one OST, the client may not be able to create files regardless of the quota available on other OSTs.</para>
+ <para>It is very important to note that the block quota is consumed per OST and the inode quota per MDS. Therefore, when the quota is consumed on one OST (resp. MDT), the client may not be able to create files regardless of the quota available on other OSTs (resp. MDTs).</para>
+ <para>Setting the quota limit below the minimal qunit size may prevent the user/group from all file creation. It is thus recommended to use soft/hard limits which are a multiple of the number of OSTs * the minimal qunit size.</para>
</note>
- <section remap="h5">
- <title>Additional information:</title>
- <para><emphasis role="bold">Grace period</emphasis> -- The period of time (in seconds) within which users are allowed to exceed their soft limit. There are four types of grace periods:</para>
- <itemizedlist>
- <listitem>
- <para> user block soft limit</para>
- </listitem>
- <listitem>
- <para> user inode soft limit</para>
- </listitem>
- <listitem>
- <para> group block soft limit</para>
- </listitem>
- <listitem>
- <para> group inode soft limit</para>
- </listitem>
- </itemizedlist>
- <para>The grace periods are applied to all users. The user block soft limit is for all users who are using a blocks quota.</para>
- <para><emphasis role="bold">Soft limit</emphasis> -- Once you are beyond the soft limit, the quota module begins to time, but you still can write block and inode. When you are always beyond the soft limit and use up your grace time, you get the same result as the hard limit. For inodes and blocks, it is the same. Usually, the soft limit MUST be less than the hard limit; if not, the quota module never triggers the timing. If the soft limit is not needed, leave it as zero (0).</para>
- <para><emphasis role="bold">Hard limit</emphasis> -- When you are beyond the hard limit, you get <literal>-EQUOTA</literal> and cannot write inode/block any more. The hard limit is the absolute limit. When a grace period is set, you can exceed the soft limit within the grace period if are under the hard limits.</para>
- <para>Lustre quota allocation is controlled by two variables, <literal>quota_bunit_sz</literal> and <literal>quota_iunit_sz</literal> referring to KBs and inodes, respectively. These values can be accessed on the MDS as <literal>/proc/fs/lustre/mds/*/quota_*</literal> and on the OST as <literal>/proc/fs/lustre/obdfilter/*/quota_*</literal>. The <literal>quota_bunit_sz </literal>and <literal>quota_iunit_sz</literal> variables are the maximum qunit values for blocks and inodes, respectively. At any time, module lquota chooses a reasonable qunit between the minimum and maximum values.</para>
- <para>The /proc values are bounded by two other variables <literal>quota_btune_sz</literal> and <literal>quota_itune_sz</literal>. By default, the <literal>*tune_sz</literal> variables are set at 1/2 the <literal>*unit_sz</literal> variables, and you cannot set <literal>*tune_sz</literal> larger than <literal>*unit_sz</literal>. You must set <literal>bunit_sz</literal> first if it is increasing by more than 2x, and <literal>btune_sz</literal> first if it is decreasing by more than 2x.</para>
- <para><emphasis role="bold">Total number of inodes</emphasis> -- To determine the total number of inodes, use <literal>lfs df -i</literal> (and also <literal>/proc/fs/lustre/*/*/filestotal</literal>). For more information on using the <literal>lfs df -i</literal> command and the command output, see <xref linkend="dbdoclet.50438209_35838"/>.</para>
- <para>Unfortunately, the <literal>statfs</literal> interface does not report the free inode count directly, but instead reports the total inode and used inode counts. The free inode count is calculated for <literal>df</literal> from (total inodes - used inodes).</para>
- <para>It is not critical to know a file system's total inode count. Instead, you should know (accurately), the free inode count and the used inode count for a file system. Lustre manipulates the total inode count in order to accurately report the other two values.</para>
- <para>The values set for the MDS must match the values set on the OSTs.</para>
- <para>The <literal>quota_bunit_sz</literal> parameter displays bytes, however <literal>lfs setquota</literal> uses KBs. The <literal>quota_bunit_sz</literal> parameter must be a multiple of 1024. A proper minimum KB size for <literal>lfs setquota</literal> can be calculated as:</para>
- <informalexample>
- <para>
- <emphasis role="bold">Size in KBs = minimum_quota_bunit_sz * (number of OSTS + 1) = 1024 * (number of OSTs +1)</emphasis>
- </para>
- </informalexample>
- <para>We add one (1) to the number of OSTs as the MDS also consumes KBs. As inodes are only consumed on the MDS, the minimum inode size for <literal>lfs setquota</literal> is equal to <literal>quota_iunit_sz</literal>.</para>
- <note>
- <para>Setting the quota below this limit may prevent the user from all file creation.</para>
- </note>
- </section>
+ <para> To determine the total number of inodes, use <literal>lfs df -i</literal> (and also <literal>lctl get_param *.*.filestotal</literal>). For more information on using the <literal>lfs df -i</literal> command and the command output, see <xref linkend="dbdoclet.50438209_35838"/>.</para>
+ <para>Unfortunately, the <literal>statfs</literal> interface does not report the free inode count directly, but instead reports the total inode and used inode counts. The free inode count is calculated for <literal>df</literal> from (total inodes - used inodes). It is not critical to know a file system's total inode count. Instead, you should know (accurately), the free inode count and the used inode count for a file system. Lustre manipulates the total inode count in order to accurately report the other two values.</para>
</section>
- <section xml:id="dbdoclet.50438217_27895">
- <title><indexterm><primary>Quotas</primary><secondary>known issues</secondary></indexterm>Known Issues with Quotas</title>
- <para>Using quotas in Lustre can be complex and there are several known issues.</para>
- <section remap="h3">
- <title>Granted Cache and Quota Limits</title>
- <para>In Lustre, granted cache does not respect quota limits. In this situation, OSTs grant cache to Lustre client to accelerate I/O. Granting cache causes writes to be successful in OSTs, even if they exceed the quota limits, and will overwrite them.</para>
- <para>The sequence is:</para>
- <orderedlist>
- <listitem>
- <para>A user writes files to Lustre.</para>
- </listitem>
- <listitem>
- <para>If the Lustre client has enough granted cache, then it returns 'success' to users and arranges the writes to the OSTs.</para>
- </listitem>
- <listitem>
- <para>Because Lustre clients have delivered success to users, the OSTs cannot fail these writes.</para>
- </listitem>
- </orderedlist>
- <para>Because of granted cache, writes always overwrite quota limitations. For example, if you set a 400 GB quota on user A and use IOR to write for user A from a bundle of clients, you will write much more data than 400 GB, and cause an out-of-quota error (<literal>-EDQUOT</literal>).</para>
- <note>
- <para>The effect of granted cache on quota limits can be mitigated, but not eradicated. Reduce the maximum amount of dirty data on the clients (can be set from 1 to 512):</para>
- <itemizedlist>
- <listitem>
- <para><literal>lctl set_param osc.*.max_dirty_mb=0</literal></para>
- </listitem>
- </itemizedlist>
- </note>
- </section>
- <section xml:id="dbdoclet.50438217_66360">
- <title><indexterm><primary>Quotas</primary><secondary>file formats</secondary></indexterm>Quota File Formats</title>
- <para>Lustre 1.6.5 introduced the v2 file format for administrative quotas, with 64-bit limits that support large-limits handling. The old quota file format (v1), with 32-bit limits, is also supported. Lustre 1.6.6 introduced the v2 file format for operational quotas. A few notes regarding the current quota file formats:</para>
- <para>From Lustre 1.6.5 until Lustre 2.0 <literal>mdt.quota_type</literal> is used to force a specific administrative quota version (v2 or v1). From Lustre 2.0 and later <literal>mdd.quota_type</literal> is used for the same effect.</para>
- <itemizedlist>
- <listitem>
- <para> For the v2 quota file format, (<literal>OBJECTS/admin_quotafile_v2.{usr,grp}</literal>)</para>
- </listitem>
- <listitem>
- <para> For the v1 quota file format, (<literal>OBJECTS/admin_quotafile.{usr,grp}</literal>)</para>
- </listitem>
- </itemizedlist>
- <para>Lustre 1.6.6 and later use <literal>ost.quota_type</literal> to force a specific operational quota version (v2 or v1).</para>
- <itemizedlist>
- <listitem>
- <para>For the v2 quota file format, (<literal>lquota_v2.{user,group}</literal>)</para>
- </listitem>
- <listitem>
- <para>For the v1 quota file format, (<literal>lquota.{user,group}</literal>)</para>
- </listitem>
- </itemizedlist>
- <para>The quota_type specifier can be used to set different combinations of administrative/operational quota file versions on a Lustre node:</para>
+ <section xml:id="quota_interoperability">
+ <title><indexterm><primary>Quotas</primary><secondary>Interoperability</secondary></indexterm>Interoperability</title>
+ <para>The new quota protocol introduced in Lustre 2.4.0 <emphasis role="bold">isn't compatible</emphasis> with the old one. As a consequence, <emphasis role="bold">all Lustre servers must be upgraded to 2.4.0 for quota to be functional</emphasis>. Quota limits set on the Lustre filesystem prior to the upgrade will be automatically migrated to the new quota index format. As for accounting information with ldiskfs backend, they will be regenerated by running <literal>tunefs.lustre --quota</literal> against all targets. It is worth noting that running <literal>tunefs.lustre --quota</literal> is <emphasis role="bold">mandatory</emphasis> for all targets formatted with a Lustre version older than 2.4.0, otherwise quota enforcement as well as accounting won't be functional.</para>
+ <para>Besides, the 2.4 quota protocol takes for granted that the Lustre client supports the <literal>OBD_CONNECT_EINPROGRESS</literal> connect flag. Clients supporting this flag will retry indefinitely when the server returns <literal>EINPROGRESS</literal> in a reply. Here is the list of Lustre client version which are compatible with 2.4:</para>
+ <itemizedlist>
+ <listitem>
+ <para>2.3-based clients and beyond</para>
+ </listitem>
+ <listitem>
+ <para>1.8 clients newer or equal to 1.8.9-wc1</para>
+ </listitem>
+ <listitem>
+ <para>2.1 clients newer or equal to 2.1.4</para>
+ </listitem>
+ </itemizedlist>
+ </section>
+ <section xml:id="granted_cache_and_quota_limits">
+ <title><indexterm><primary>Quotas</primary><secondary>known issues</secondary></indexterm>Granted Cache and Quota Limits</title>
+ <para>In Lustre, granted cache does not respect quota limits. In this situation, OSTs grant cache to Lustre client to accelerate I/O. Granting cache causes writes to be successful in OSTs, even if they exceed the quota limits, and will overwrite them.</para>
+ <para>The sequence is:</para>
+ <orderedlist>
+ <listitem>
+ <para>A user writes files to Lustre.</para>
+ </listitem>
+ <listitem>
+ <para>If the Lustre client has enough granted cache, then it returns 'success' to users and arranges the writes to the OSTs.</para>
+ </listitem>
+ <listitem>
+ <para>Because Lustre clients have delivered success to users, the OSTs cannot fail these writes.</para>
+ </listitem>
+ </orderedlist>
+ <para>Because of granted cache, writes always overwrite quota limitations. For example, if you set a 400 GB quota on user A and use IOR to write for user A from a bundle of clients, you will write much more data than 400 GB, and cause an out-of-quota error (<literal>EDQUOT</literal>).</para>
+ <note>
+ <para>The effect of granted cache on quota limits can be mitigated, but not eradicated. Reduce the maximum amount of dirty data on the clients (minimal value is 1MB):</para>
<itemizedlist>
<listitem>
- <para>"1" - v1 (32-bit) administrative quota file, v1 (32-bit) operational quota file (default in releases <emphasis role="underline">before</emphasis> Lustre 1.6.5)</para>
- </listitem>
- <listitem>
- <para>"2" - v2 (64-bit) administrative quota file, v1 (32-bit) operational quota file (default in Lustre 1.6.5)</para>
- </listitem>
- <listitem>
- <para>"3" - v2 (64-bit) administrative quota file, v2 (64-bit) operational quota file (default in releases <emphasis role="underline">after</emphasis> Lustre 1.6.5)</para>
+ <para><literal>lctl set_param osc.*.max_dirty_mb=8</literal></para>
</listitem>
</itemizedlist>
- <para>If quotas do not exist or look broken, then <literal>quotacheck</literal> creates quota files of a required name and format.</para>
- <para>If Lustre is using the v2 quota file format when only v1 quota files exist, then <literal>quotacheck</literal> converts old v1 quota files to new v2 quota files. This conversion is triggered automatically, and is transparent to users. If an old quota file does not exist or looks broken, then the new v2 quota file will be empty. In case of an error, details can be found in the kernel log of the corresponding MDS/OST. During conversion of a v1 quota file to a v2 quota file, the v2 quota file is marked as broken, to avoid it being used if a crash occurs. The quota module does not use broken quota files (keeping quota off).</para>
- <para>In most situations, Lustre administrators do not need to set specific versioning options. Upgrading Lustre without using <literal>quota_type</literal> to force specific quota file versions results in quota files being upgraded automatically to the latest version. The option ensures backward compatibility, preventing a quota file upgrade to a version which is not supported by earlier Lustre versions.</para>
- </section>
+ </note>
</section>
- <section xml:id="dbdoclet.50438217_20772">
+ <section xml:id="lustre_quota_statistics">
<title><indexterm><primary>Quotas</primary><secondary>statistics</secondary></indexterm>Lustre Quota Statistics</title>
<para>Lustre includes statistics that monitor quota activity, such as the kinds of quota RPCs sent during a specific period, the average time to complete the RPCs, etc. These statistics are useful to measure performance of a Lustre file system.</para>
<para>Each quota statistic consists of a quota event and <literal>min_time</literal>, <literal>max_time</literal> and <literal>sum_time</literal> values for the event.</para>
<title>Interpreting Quota Statistics</title>
<para>Quota statistics are an important measure of a Lustre file system's performance. Interpreting these statistics correctly can help you diagnose problems with quotas, and may indicate adjustments to improve system performance.</para>
<para>For example, if you run this command on the OSTs:</para>
- <screen>cat /proc/fs/lustre/lquota/lustre-OST0000/stats</screen>
+ <screen>lctl get_param lquota.testfs-OST0000.stats</screen>
<para>You will get a result similar to this:</para>
<screen>snapshot_time 1219908615.506895 secs.usecs
async_acq_req 1 samples [us] 32 32 32
git://git.whamcloud.com / doc / manual.git / commitdiff