1 <?xml version='1.0' encoding='UTF-8'?><chapter xmlns="http://docbook.org/ns/docbook" xmlns:xl="http://www.w3.org/1999/xlink" version="5.0" xml:lang="en-US" xml:id="lustremonitoring">
2 <title xml:id="lustremonitoring.title">Monitoring a Lustre File System</title>
3 <para>This chapter provides information on monitoring a Lustre file system and includes the
4 following sections:</para>
7 <para><xref linkend="dbdoclet.50438273_18711"/>Lustre Changelogs</para>
10 <para><xref linkend="dbdoclet.jobstats"/>Lustre Jobstats</para>
13 <para><xref linkend="dbdoclet.50438273_81684"/>Lustre Monitoring Tool</para>
16 <para><xref linkend="dbdoclet.50438273_80593"/>CollectL</para>
19 <para><xref linkend="dbdoclet.50438273_44185"/>Other Monitoring Options</para>
22 <section xml:id="dbdoclet.50438273_18711">
23 <title><indexterm><primary>change logs</primary><see>monitoring</see></indexterm>
24 <indexterm><primary>monitoring</primary></indexterm>
25 <indexterm><primary>monitoring</primary><secondary>change logs</secondary></indexterm>
27 Lustre Changelogs</title>
28 <para>The changelogs feature records events that change the file system
29 namespace or file metadata. Changes such as file creation, deletion,
30 renaming, attribute changes, etc. are recorded with the target and parent
31 file identifiers (FIDs), the name of the target, a timestamp, and user
32 information. These records can be used for a variety of purposes:</para>
35 <para>Capture recent changes to feed into an archiving system.</para>
38 <para>Use changelog entries to exactly replicate changes in a file
42 <para>Set up "watch scripts" that take action on certain
43 events or directories.</para>
46 <para>Audit activity on Lustre, thanks to user information associated to
47 file/directory changes with timestamps.</para>
50 <para>Changelogs record types are:</para>
51 <informaltable frame="all">
53 <colspec colname="c1" colwidth="50*"/>
54 <colspec colname="c2" colwidth="50*"/>
58 <para><emphasis role="bold">Value</emphasis></para>
61 <para><emphasis role="bold">Description</emphasis></para>
71 <para> Internal recordkeeping</para>
79 <para> Regular file creation</para>
87 <para> Directory creation</para>
95 <para> Hard link</para>
103 <para> Soft link</para>
111 <para> Other file creation</para>
119 <para> Regular file removal</para>
127 <para> Directory removal</para>
135 <para> Rename, original</para>
143 <para> Rename, final</para>
167 <para> Layout change</para>
175 <para> Regular file truncated</para>
183 <para> Attribute change</para>
191 <para> Extended attribute change (setxattr)</para>
199 <para> HSM specific event</para>
207 <para> MTIME change</para>
215 <para> CTIME change</para>
220 <para> ATIME *</para>
223 <para> ATIME change</para>
231 <para> Migration event</para>
239 <para> File Level Replication: file initially written</para>
247 <para> File Level Replication: file re-synced</para>
252 <para> GXATR *</para>
255 <para> Extended attribute access (getxattr)</para>
260 <para> NOPEN *</para>
263 <para> Denied open</para>
269 <note><para>Event types marked with * are not recorded by default. Refer to
270 <xref linkend="dbdoclet.modifyChangelogMask" /> for instructions on
271 modifying the Changelogs mask.</para></note>
272 <para>FID-to-full-pathname and pathname-to-FID functions are also included
273 to map target and parent FIDs into the file system namespace.</para>
275 <title><indexterm><primary>monitoring</primary><secondary>change logs
276 </secondary></indexterm>
277 Working with Changelogs</title>
278 <para>Several commands are available to work with changelogs.</para>
281 <literal>lctl changelog_register</literal>
283 <para>Because changelog records take up space on the MDT, the system
284 administration must register changelog users. As soon as a changelog
285 user is registered, the Changelogs feature is enabled. The registrants
286 specify which records they are "done with", and the system
287 purges up to the greatest common record.</para>
288 <para>To register a new changelog user, run:</para>
289 <screen>mds# lctl --device <replaceable>fsname</replaceable>-<replaceable>MDTnumber</replaceable> changelog_register
291 <para>Changelog entries are not purged beyond a registered user's
292 set point (see <literal>lfs changelog_clear</literal>).</para>
296 <literal>lfs changelog</literal>
298 <para>To display the metadata changes on an MDT (the changelog records),
300 <screen>lfs changelog <replaceable>fsname</replaceable>-<replaceable>MDTnumber</replaceable> [startrec [endrec]] </screen>
301 <para>It is optional whether to specify the start and end
303 <para>These are sample changelog records:</para>
304 <screen>1 02MKDIR 15:15:21.977666834 2018.01.09 0x0 t=[0x200000402:0x1:0x0] j=mkdir.500 ef=0xf \
305 u=500:500 nid=10.128.11.159@tcp p=[0x200000007:0x1:0x0] pics
306 2 01CREAT 15:15:36.687592024 2018.01.09 0x0 t=[0x200000402:0x2:0x0] j=cp.500 ef=0xf \
307 u=500:500 nid=10.128.11.159@tcp p=[0x200000402:0x1:0x0] chloe.jpg
308 3 06UNLNK 15:15:41.305116815 2018.01.09 0x1 t=[0x200000402:0x2:0x0] j=rm.500 ef=0xf \
309 u=500:500 nid=10.128.11.159@tcp p=[0x200000402:0x1:0x0] chloe.jpg
310 4 07RMDIR 15:15:46.468790091 2018.01.09 0x1 t=[0x200000402:0x1:0x0] j=rmdir.500 ef=0xf \
311 u=500:500 nid=10.128.11.159@tcp p=[0x200000007:0x1:0x0] pics </screen>
315 <literal>lfs changelog_clear</literal>
317 <para>To clear old changelog records for a specific user (records that
318 the user no longer needs), run:</para>
319 <screen>lfs changelog_clear <replaceable>mdt_name</replaceable> <replaceable>userid</replaceable> <replaceable>endrec</replaceable></screen>
320 <para>The <literal>changelog_clear</literal> command indicates that
321 changelog records previous to <replaceable>endrec</replaceable> are no
322 longer of interest to a particular user
323 <replaceable>userid</replaceable>, potentially allowing the MDT to free
324 up disk space. An <literal><replaceable>endrec</replaceable></literal>
325 value of 0 indicates the current last record. To run
326 <literal>changelog_clear</literal>, the changelog user must be
327 registered on the MDT node using <literal>lctl</literal>.</para>
328 <para>When all changelog users are done with records < X, the records
333 <literal>lctl changelog_deregister</literal>
335 <para>To deregister (unregister) a changelog user, run:</para>
336 <screen>mds# lctl --device <replaceable>mdt_device</replaceable> changelog_deregister <replaceable>userid</replaceable> </screen>
337 <para> <literal>changelog_deregister cl1</literal> effectively does a
338 <literal>lfs changelog_clear cl1 0</literal> as it deregisters.</para>
342 <title>Changelog Examples</title>
343 <para>This section provides examples of different changelog
346 <title>Registering a Changelog User</title>
347 <para>To register a new changelog user for a device
348 (<literal>lustre-MDT0000</literal>):</para>
349 <screen>mds# lctl --device lustre-MDT0000 changelog_register
350 lustre-MDT0000: Registered changelog userid 'cl1'</screen>
353 <title>Displaying Changelog Records</title>
354 <para>To display changelog records on an MDT
355 (<literal>lustre-MDT0000</literal>):</para>
356 <screen>$ lfs changelog lustre-MDT0000
357 1 02MKDIR 15:15:21.977666834 2018.01.09 0x0 t=[0x200000402:0x1:0x0] ef=0xf \
358 u=500:500 nid=10.128.11.159@tcp p=[0x200000007:0x1:0x0] pics
359 2 01CREAT 15:15:36.687592024 2018.01.09 0x0 t=[0x200000402:0x2:0x0] ef=0xf \
360 u=500:500 nid=10.128.11.159@tcp p=[0x200000402:0x1:0x0] chloe.jpg
361 3 06UNLNK 15:15:41.305116815 2018.01.09 0x1 t=[0x200000402:0x2:0x0] ef=0xf \
362 u=500:500 nid=10.128.11.159@tcp p=[0x200000402:0x1:0x0] chloe.jpg
363 4 07RMDIR 15:15:46.468790091 2018.01.09 0x1 t=[0x200000402:0x1:0x0] ef=0xf \
364 u=500:500 nid=10.128.11.159@tcp p=[0x200000007:0x1:0x0] pics</screen>
365 <para>Changelog records include this information:</para>
367 operation_type(numerical/text)
377 <para>Displayed in this format:</para>
378 <screen>rec# operation_type(numerical/text) timestamp datestamp flags t=target_FID \
379 ef=extended_flags u=uid:gid nid=client_NID p=parent_FID target_name</screen>
380 <para>For example:</para>
381 <screen>2 01CREAT 15:15:36.687592024 2018.01.09 0x0 t=[0x200000402:0x2:0x0] ef=0xf \
382 u=500:500 nid=10.128.11.159@tcp p=[0x200000402:0x1:0x0] chloe.jpg</screen>
385 <title>Clearing Changelog Records</title>
386 <para>To notify a device that a specific user (<literal>cl1</literal>)
387 no longer needs records (up to and including 3):</para>
388 <screen>$ lfs changelog_clear lustre-MDT0000 cl1 3</screen>
389 <para>To confirm that the <literal>changelog_clear</literal> operation
390 was successful, run <literal>lfs changelog</literal>; only records after
391 id-3 are listed:</para>
392 <screen>$ lfs changelog lustre-MDT0000
393 4 07RMDIR 15:15:46.468790091 2018.01.09 0x1 t=[0x200000402:0x1:0x0] ef=0xf \
394 u=500:500 nid=10.128.11.159@tcp p=[0x200000007:0x1:0x0] pics</screen>
397 <title>Deregistering a Changelog User</title>
398 <para>To deregister a changelog user (<literal>cl1</literal>) for a
399 specific device (<literal>lustre-MDT0000</literal>):</para>
400 <screen>mds# lctl --device lustre-MDT0000 changelog_deregister cl1
401 lustre-MDT0000: Deregistered changelog user 'cl1'</screen>
402 <para>The deregistration operation clears all changelog records for the
403 specified user (<literal>cl1</literal>).</para>
404 <screen>$ lfs changelog lustre-MDT0000
405 5 00MARK 15:56:39.603643887 2018.01.09 0x0 t=[0x20001:0x0:0x0] ef=0xf \
406 u=500:500 nid=0@<0:0> p=[0:0x50:0xb] mdd_obd-lustre-MDT0000-0
409 <para>MARK records typically indicate changelog recording status
414 <title>Displaying the Changelog Index and Registered Users</title>
415 <para>To display the current, maximum changelog index and registered
416 changelog users for a specific device
417 (<literal>lustre-MDT0000</literal>):</para>
418 <screen>mds# lctl get_param mdd.lustre-MDT0000.changelog_users
419 mdd.lustre-MDT0000.changelog_users=current index: 8
420 ID index (idle seconds)
425 <title>Displaying the Changelog Mask</title>
426 <para>To show the current changelog mask on a specific device
427 (<literal>lustre-MDT0000</literal>):</para>
428 <screen>mds# lctl get_param mdd.lustre-MDT0000.changelog_mask
430 mdd.lustre-MDT0000.changelog_mask=
431 MARK CREAT MKDIR HLINK SLINK MKNOD UNLNK RMDIR RENME RNMTO CLOSE LYOUT \
432 TRUNC SATTR XATTR HSM MTIME CTIME MIGRT
435 <section xml:id="dbdoclet.modifyChangelogMask" remap="h5">
436 <title>Setting the Changelog Mask</title>
437 <para>To set the current changelog mask on a specific device
438 (<literal>lustre-MDT0000</literal>):</para>
439 <screen>mds# lctl set_param mdd.lustre-MDT0000.changelog_mask=HLINK
440 mdd.lustre-MDT0000.changelog_mask=HLINK
441 $ lfs changelog_clear lustre-MDT0000 cl1 0
442 $ mkdir /mnt/lustre/mydir/foo
443 $ cp /etc/hosts /mnt/lustre/mydir/foo/file
444 $ ln /mnt/lustre/mydir/foo/file /mnt/lustre/mydir/myhardlink
446 <para>Only item types that are in the mask show up in the
448 <screen>$ lfs changelog lustre-MDT0000
449 9 03HLINK 16:06:35.291636498 2018.01.09 0x0 t=[0x200000402:0x4:0x0] ef=0xf \
450 u=500:500 nid=10.128.11.159@tcp p=[0x200000007:0x3:0x0] myhardlink
455 <section remap="h3" condition='l2B'>
456 <title><indexterm><primary>audit</primary>
457 <secondary>change logs</secondary></indexterm>
458 Audit with Changelogs</title>
459 <para>A specific use case for Lustre Changelogs is audit. According to a
460 definition found on <link xmlns:xlink="http://www.w3.org/1999/xlink"
461 xlink:href="https://en.wikipedia.org/wiki/Information_technology_audit">
462 Wikipedia</link>, information technology audits are used to evaluate the
463 organization's ability to protect its information assets and to properly
464 dispense information to authorized parties. Basically, audit consists in
465 controlling that all data accesses made were done according to the access
466 control policy in place. And usually, this is done by analyzing access
468 <para>Audit can be used as a proof of security in place. But Audit can
469 also be a requirement to comply with regulations.</para>
470 <para>Lustre Changelogs are a good mechanism for audit, because this is a
471 centralized facility, and it is designed to be transactional. Changelog
472 records contain all information necessary for auditing purposes:</para>
475 <para>ability to identify object of action thanks to file identifiers
476 (FIDs) and name of targets</para>
479 <para>ability to identify subject of action thanks to UID/GID and NID
483 <para>ability to identify time of action thanks to timestamp</para>
487 <title>Enabling Audit</title>
488 <para>To have a fully functional Changelogs-based audit facility, some
489 additional Changelog record types must be enabled, to be able to record
490 events such as OPEN, ATIME, GETXATTR and DENIED OPEN. Please note that
491 enabling these record types may have some performance impact. For
492 instance, recording OPEN and GETXATTR events generate writes in the
493 Changelog records for a read operation from a file-system
495 <para>Being able to record events such as OPEN or DENIED OPEN is
496 important from an audit perspective. For instance, if Lustre file system
497 is used to store medical records on a system dedicated to Life Sciences,
498 data privacy is crucial. Administrators may need to know which doctors
499 accessed, or tried to access, a given medical record and when. And
500 conversely, they might need to know which medical records a given doctor
502 <para>To enable all changelog entry types, do:</para>
503 <screen>mds# lctl set_param mdd.lustre-MDT0000.changelog_mask=ALL
504 mdd.seb-MDT0000.changelog_mask=ALL</screen>
505 <para>Once all required record types have been enabled, just register a
506 Changelogs user and the audit facility is operational.</para>
507 <para>Note that, however, it is possible to control which Lustre client
508 nodes can trigger the recording of file system access events to the
509 Changelogs, thanks to the <literal>audit_mode</literal> flag on nodemap
510 entries. The reason to disable audit on a per-nodemap basis is to
511 prevent some nodes (e.g. backup, HSM agent nodes) from flooding the
512 audit logs. When <literal>audit_mode</literal> flag is
513 set to 1 on a nodemap entry, a client pertaining to this nodemap will be
514 able to record file system access events to the Changelogs, if
515 Changelogs are otherwise activated. When set to 0, events are not logged
516 into the Changelogs, no matter if Changelogs are activated or not. By
517 default, <literal>audit_mode</literal> flag is set to 1 in newly created
518 nodemap entries. And it is also set to 1 in 'default' nodemap.</para>
519 <para>To prevent nodes pertaining to a nodemap to generate Changelog
522 mgs# lctl nodemap_modify --name nm1 --property audit_mode --value 0</screen>
525 <title>Audit examples</title>
528 <literal>OPEN</literal>
530 <para>An OPEN changelog entry is in the form:</para>
532 7 10OPEN 13:38:51.510728296 2017.07.25 0x242 t=[0x200000401:0x2:0x0] \
533 ef=0x7 u=500:500 nid=10.128.11.159@tcp m=-w-</screen>
534 <para>It includes information about the open mode, in the form
536 <para>OPEN entries are recorded only once per UID/GID, for a given
537 open mode, as long as the file is not closed by this UID/GID. It
538 avoids flooding the Changelogs for instance if there is an MPI job
539 opening the same file thousands of times from different threads. It
540 reduces the ChangeLog load significantly, without significantly
541 affecting the audit information. Similarly, only the last CLOSE per
542 UID/GID is recorded.</para>
546 <literal>GETXATTR</literal>
548 <para>A GETXATTR changelog entry is in the form:</para>
550 8 23GXATR 09:22:55.886793012 2017.07.27 0x0 t=[0x200000402:0x1:0x0] \
551 ef=0xf u=500:500 nid=10.128.11.159@tcp x=user.name0</screen>
552 <para>It includes information about the name of the extended attribute
553 being accessed, in the form <literal>x=<xattr name></literal>.
558 <literal>SETXATTR</literal>
560 <para>A SETXATTR changelog entry is in the form:</para>
562 4 15XATTR 09:41:36.157333594 2018.01.10 0x0 t=[0x200000402:0x1:0x0] \
563 ef=0xf u=500:500 nid=10.128.11.159@tcp x=user.name0</screen>
564 <para>It includes information about the name of the extended attribute
565 being modified, in the form <literal>x=<xattr name></literal>.
570 <literal>DENIED OPEN</literal>
572 <para>A DENIED OPEN changelog entry is in the form:</para>
574 4 24NOPEN 15:45:44.947406626 2017.08.31 0x2 t=[0x200000402:0x1:0x0] \
575 ef=0xf u=500:500 nid=10.128.11.158@tcp m=-w-</screen>
576 <para>It has the same information as a regular OPEN entry. In order to
577 avoid flooding the Changelogs, DENIED OPEN entries are rate limited:
578 no more than one entry per user per file per time interval, this time
579 interval (in seconds) being configurable via
580 <literal>mdd.<mdtname>.changelog_deniednext</literal>
581 (default value is 60 seconds).</para>
583 mds# lctl set_param mdd.lustre-MDT0000.changelog_deniednext=120
584 mdd.seb-MDT0000.changelog_deniednext=120
585 mds# lctl get_param mdd.lustre-MDT0000.changelog_deniednext
586 mdd.seb-MDT0000.changelog_deniednext=120</screen>
591 <section xml:id="dbdoclet.jobstats">
592 <title><indexterm><primary>jobstats</primary><see>monitoring</see></indexterm>
593 <indexterm><primary>monitoring</primary></indexterm>
594 <indexterm><primary>monitoring</primary><secondary>jobstats</secondary></indexterm>
596 Lustre Jobstats</title>
597 <para>The Lustre jobstats feature collects file system operation statistics
598 for user processes running on Lustre clients, and exposes on the server
599 using the unique Job Identifier (JobID) provided by the job scheduler for
600 each job. Job schedulers known to be able to work with jobstats include:
601 SLURM, SGE, LSF, Loadleveler, PBS and Maui/MOAB.</para>
602 <para>Since jobstats is implemented in a scheduler-agnostic manner, it is
603 likely that it will be able to work with other schedulers also, and also
604 in environments that do not use a job scheduler, by storing custom format
605 strings in the <literal>jobid_name</literal>.</para>
607 <title><indexterm><primary>monitoring</primary><secondary>jobstats</secondary></indexterm>
608 How Jobstats Works</title>
609 <para>The Lustre jobstats code on the client extracts the unique JobID
610 from an environment variable within the user process, and sends this
611 JobID to the server with the I/O operation. The server tracks
612 statistics for operations whose JobID is given, indexed by that
615 <para>A Lustre setting on the client, <literal>jobid_var</literal>,
616 specifies which environment variable to holds the JobID for that process
617 Any environment variable can be specified. For example, SLURM sets the
618 <literal>SLURM_JOB_ID</literal> environment variable with the unique
619 job ID on each client when the job is first launched on a node, and
620 the <literal>SLURM_JOB_ID</literal> will be inherited by all child
621 processes started below that process.</para>
623 <para>Lustre can be configured to generate a synthetic JobID from
624 the client's process name and numeric UID, by setting
625 <literal>jobid_var=procname_uid</literal>. This will generate a
626 uniform JobID when running the same binary across multiple client
627 nodes, but cannot distinguish whether the binary is part of a single
628 distributed process or multiple independent processes.
631 <para condition="l28">In Lustre 2.8 and later it is possible to set
632 <literal>jobid_var=nodelocal</literal> and then also set
633 <literal>jobid_name=</literal><replaceable>name</replaceable>, which
634 <emphasis>all</emphasis> processes on that client node will use. This
635 is useful if only a single job is run on a client at one time, but if
636 multiple jobs are run on a client concurrently, the per-session JobID
640 <para condition="l2C">In Lustre 2.12 and later, it is possible to
641 specify more complex JobID values for <literal>jobid_name</literal>
642 by using a string that contains format codes that are evaluated for
643 each process, in order to generate a site- or node-specific JobID string.
647 <para><emphasis>%e</emphasis> print executable name</para>
650 <para><emphasis>%g</emphasis> print group ID number</para>
653 <para><emphasis>%h</emphasis> print hostname</para>
656 <para><emphasis>%j</emphasis> print JobID from process environment
657 variable named by the <emphasis>jobid_var</emphasis> parameter
661 <para><emphasis>%p</emphasis> print numeric process ID</para>
664 <para><emphasis>%u</emphasis> print user ID number</para>
668 <para condition="l2D">In Lustre 2.13 and later, it is possible to
669 set a per-session JobID by setting the
670 <literal>jobid_this_session</literal> parameter. This will be
671 inherited by all processes that are started in this login session,
672 but there can be a different JobID for each login session.
675 <para>The setting of <literal>jobid_var</literal> need not be the same
676 on all clients. For example, one could use
677 <literal>SLURM_JOB_ID</literal> on all clients managed by SLURM, and
678 use <literal>procname_uid</literal> on clients not managed by SLURM,
679 such as interactive login nodes.</para>
681 <para>It is not possible to have different
682 <literal>jobid_var</literal> settings on a single node, since it is
683 unlikely that multiple job schedulers are active on one client.
684 However, the actual JobID value is local to each process environment
685 and it is possible for multiple jobs with different JobIDs to be
686 active on a single client at one time.</para>
690 <title><indexterm><primary>monitoring</primary><secondary>jobstats</secondary></indexterm>
691 Enable/Disable Jobstats</title>
692 <para>Jobstats are disabled by default. The current state of jobstats
693 can be verified by checking <literal>lctl get_param jobid_var</literal>
696 $ lctl get_param jobid_var
700 To enable jobstats on the <literal>testfs</literal> file system with SLURM:</para>
701 <screen># lctl conf_param testfs.sys.jobid_var=SLURM_JOB_ID</screen>
702 <para>The <literal>lctl conf_param</literal> command to enable or disable
703 jobstats should be run on the MGS as root. The change is persistent, and
704 will be propagated to the MDS, OSS, and client nodes automatically when
705 it is set on the MGS and for each new client mount.</para>
706 <para>To temporarily enable jobstats on a client, or to use a different
707 jobid_var on a subset of nodes, such as nodes in a remote cluster that
708 use a different job scheduler, or interactive login nodes that do not
709 use a job scheduler at all, run the <literal>lctl set_param</literal>
710 command directly on the client node(s) after the filesystem is mounted.
711 For example, to enable the <literal>procname_uid</literal> synthetic
712 JobID on a login node run:
713 <screen># lctl set_param jobid_var=procname_uid</screen>
714 The <literal>lctl set_param</literal> setting is not persistent, and will
715 be reset if the global <literal>jobid_var</literal> is set on the MGS or
716 if the filesystem is unmounted.</para>
717 <para>The following table shows the environment variables which are set
718 by various job schedulers. Set <literal>jobid_var</literal> to the value
719 for your job scheduler to collect statistics on a per job basis.</para>
720 <informaltable frame="all">
722 <colspec colname="c1" colwidth="50*"/>
723 <colspec colname="c2" colwidth="50*"/>
727 <para><emphasis role="bold">Job Scheduler</emphasis></para>
730 <para><emphasis role="bold">Environment Variable</emphasis></para>
737 <para>Simple Linux Utility for Resource Management (SLURM)</para>
740 <para>SLURM_JOB_ID</para>
745 <para>Sun Grid Engine (SGE)</para>
753 <para>Load Sharing Facility (LSF)</para>
756 <para>LSB_JOBID</para>
761 <para>Loadleveler</para>
764 <para>LOADL_STEP_ID</para>
769 <para>Portable Batch Scheduler (PBS)/MAUI</para>
772 <para>PBS_JOBID</para>
777 <para>Cray Application Level Placement Scheduler (ALPS)</para>
780 <para>ALPS_APP_ID</para>
786 <para>There are two special values for <literal>jobid_var</literal>:
787 <literal>disable</literal> and <literal>procname_uid</literal>. To disable
788 jobstats, specify <literal>jobid_var</literal> as <literal>disable</literal>:</para>
789 <screen># lctl conf_param testfs.sys.jobid_var=disable</screen>
790 <para>To track job stats per process name and user ID (for debugging, or
791 if no job scheduler is in use on some nodes such as login nodes), specify
792 <literal>jobid_var</literal> as <literal>procname_uid</literal>:</para>
793 <screen># lctl conf_param testfs.sys.jobid_var=procname_uid</screen>
796 <title><indexterm><primary>monitoring</primary><secondary>jobstats</secondary></indexterm>
797 Check Job Stats</title>
798 <para>Metadata operation statistics are collected on MDTs. These statistics can be accessed for
799 all file systems and all jobs on the MDT via the <literal>lctl get_param
800 mdt.*.job_stats</literal>. For example, clients running with
801 <literal>jobid_var=procname_uid</literal>:</para>
803 # lctl get_param mdt.*.job_stats
806 snapshot_time: 1352084992
807 open: { samples: 2, unit: reqs }
808 close: { samples: 2, unit: reqs }
809 mknod: { samples: 0, unit: reqs }
810 link: { samples: 0, unit: reqs }
811 unlink: { samples: 0, unit: reqs }
812 mkdir: { samples: 0, unit: reqs }
813 rmdir: { samples: 0, unit: reqs }
814 rename: { samples: 0, unit: reqs }
815 getattr: { samples: 3, unit: reqs }
816 setattr: { samples: 0, unit: reqs }
817 getxattr: { samples: 0, unit: reqs }
818 setxattr: { samples: 0, unit: reqs }
819 statfs: { samples: 0, unit: reqs }
820 sync: { samples: 0, unit: reqs }
821 samedir_rename: { samples: 0, unit: reqs }
822 crossdir_rename: { samples: 0, unit: reqs }
823 - job_id: mythbackend.0
824 snapshot_time: 1352084996
825 open: { samples: 72, unit: reqs }
826 close: { samples: 73, unit: reqs }
827 mknod: { samples: 0, unit: reqs }
828 link: { samples: 0, unit: reqs }
829 unlink: { samples: 22, unit: reqs }
830 mkdir: { samples: 0, unit: reqs }
831 rmdir: { samples: 0, unit: reqs }
832 rename: { samples: 0, unit: reqs }
833 getattr: { samples: 778, unit: reqs }
834 setattr: { samples: 22, unit: reqs }
835 getxattr: { samples: 0, unit: reqs }
836 setxattr: { samples: 0, unit: reqs }
837 statfs: { samples: 19840, unit: reqs }
838 sync: { samples: 33190, unit: reqs }
839 samedir_rename: { samples: 0, unit: reqs }
840 crossdir_rename: { samples: 0, unit: reqs }
842 <para>Data operation statistics are collected on OSTs. Data operations
843 statistics can be accessed via
844 <literal>lctl get_param obdfilter.*.job_stats</literal>, for example:</para>
846 $ lctl get_param obdfilter.*.job_stats
847 obdfilter.myth-OST0000.job_stats=
849 - job_id: mythcommflag.0
850 snapshot_time: 1429714922
851 read: { samples: 974, unit: bytes, min: 4096, max: 1048576, sum: 91530035 }
852 write: { samples: 0, unit: bytes, min: 0, max: 0, sum: 0 }
853 setattr: { samples: 0, unit: reqs }
854 punch: { samples: 0, unit: reqs }
855 sync: { samples: 0, unit: reqs }
856 obdfilter.myth-OST0001.job_stats=
858 - job_id: mythbackend.0
859 snapshot_time: 1429715270
860 read: { samples: 0, unit: bytes, min: 0, max: 0, sum: 0 }
861 write: { samples: 1, unit: bytes, min: 96899, max: 96899, sum: 96899 }
862 setattr: { samples: 0, unit: reqs }
863 punch: { samples: 1, unit: reqs }
864 sync: { samples: 0, unit: reqs }
865 obdfilter.myth-OST0002.job_stats=job_stats:
866 obdfilter.myth-OST0003.job_stats=job_stats:
867 obdfilter.myth-OST0004.job_stats=
869 - job_id: mythfrontend.500
870 snapshot_time: 1429692083
871 read: { samples: 9, unit: bytes, min: 16384, max: 1048576, sum: 4444160 }
872 write: { samples: 0, unit: bytes, min: 0, max: 0, sum: 0 }
873 setattr: { samples: 0, unit: reqs }
874 punch: { samples: 0, unit: reqs }
875 sync: { samples: 0, unit: reqs }
876 - job_id: mythbackend.500
877 snapshot_time: 1429692129
878 read: { samples: 0, unit: bytes, min: 0, max: 0, sum: 0 }
879 write: { samples: 1, unit: bytes, min: 56231, max: 56231, sum: 56231 }
880 setattr: { samples: 0, unit: reqs }
881 punch: { samples: 1, unit: reqs }
882 sync: { samples: 0, unit: reqs }
886 <title><indexterm><primary>monitoring</primary><secondary>jobstats</secondary></indexterm>
887 Clear Job Stats</title>
888 <para>Accumulated job statistics can be reset by writing proc file <literal>job_stats</literal>.</para>
889 <para>Clear statistics for all jobs on the local node:</para>
890 <screen># lctl set_param obdfilter.*.job_stats=clear</screen>
891 <para>Clear statistics only for job 'bash.0' on lustre-MDT0000:</para>
892 <screen># lctl set_param mdt.lustre-MDT0000.job_stats=bash.0</screen>
895 <title><indexterm><primary>monitoring</primary><secondary>jobstats</secondary></indexterm>
896 Configure Auto-cleanup Interval</title>
897 <para>By default, if a job is inactive for 600 seconds (10 minutes) statistics for this job will be dropped. This expiration value can be changed temporarily via:</para>
898 <screen># lctl set_param *.*.job_cleanup_interval={max_age}</screen>
899 <para>It can also be changed permanently, for example to 700 seconds via:</para>
900 <screen># lctl conf_param testfs.mdt.job_cleanup_interval=700</screen>
901 <para>The <literal>job_cleanup_interval</literal> can be set as 0 to disable the auto-cleanup. Note that if auto-cleanup of Jobstats is disabled, then all statistics will be kept in memory forever, which may eventually consume all memory on the servers. In this case, any monitoring tool should explicitly clear individual job statistics as they are processed, as shown above.</para>
904 <section xml:id="dbdoclet.50438273_81684">
906 <primary>monitoring</primary>
907 <secondary>Lustre Monitoring Tool</secondary>
908 </indexterm> Lustre Monitoring Tool (LMT)</title>
909 <para>The Lustre Monitoring Tool (LMT) is a Python-based, distributed system that provides a
910 <literal>top</literal>-like display of activity on server-side nodes (MDS, OSS and portals
911 routers) on one or more Lustre file systems. It does not provide support for monitoring
912 clients. For more information on LMT, including the setup procedure, see:</para>
913 <para><link xl:href="http://code.google.com/p/lmt/"
914 >https://github.com/chaos/lmt/wiki</link></para>
915 <para>LMT questions can be directed to:</para>
916 <para><link xl:href="mailto:lmt-discuss@googlegroups.com">lmt-discuss@googlegroups.com</link></para>
918 <section xml:id="dbdoclet.50438273_80593">
920 <literal>CollectL</literal>
922 <para><literal>CollectL</literal> is another tool that can be used to monitor a Lustre file
923 system. You can run <literal>CollectL</literal> on a Lustre system that has any combination of
924 MDSs, OSTs and clients. The collected data can be written to a file for continuous logging and
925 played back at a later time. It can also be converted to a format suitable for
927 <para>For more information about <literal>CollectL</literal>, see:</para>
928 <para><link xl:href="http://collectl.sourceforge.net">http://collectl.sourceforge.net</link></para>
929 <para>Lustre-specific documentation is also available. See:</para>
930 <para><link xl:href="http://collectl.sourceforge.net/Tutorial-Lustre.html">http://collectl.sourceforge.net/Tutorial-Lustre.html</link></para>
932 <section xml:id="dbdoclet.50438273_44185">
933 <title><indexterm><primary>monitoring</primary><secondary>additional tools</secondary></indexterm>
934 Other Monitoring Options</title>
935 <para>A variety of standard tools are available publicly including the following:<itemizedlist>
937 <para><literal>lltop</literal> - Lustre load monitor with batch scheduler integration.
938 <link xmlns:xlink="http://www.w3.org/1999/xlink"
939 xlink:href="https://github.com/jhammond/lltop"
940 >https://github.com/jhammond/lltop</link></para>
943 <para><literal>tacc_stats</literal> - A job-oriented system monitor, analyzation, and
944 visualization tool that probes Lustre interfaces and collects statistics. <link
945 xmlns:xlink="http://www.w3.org/1999/xlink"
946 xlink:href="https://github.com/jhammond/tacc_stats"/></para>
949 <para><literal>xltop</literal> - A continuous Lustre monitor with batch scheduler
950 integration. <link xmlns:xlink="http://www.w3.org/1999/xlink"
951 xlink:href="https://github.com/jhammond/xltop"/></para>
953 </itemizedlist></para>
954 <para>Another option is to script a simple monitoring solution that looks at various reports
955 from <literal>ipconfig</literal>, as well as the <literal>procfs</literal> files generated by
956 the Lustre software.</para>