sreport - Man Page

• Show pipe delimited cluster utilization data: sreport --parsable cluster utilization
• Show number of jobs run: sreport job sizes printjobcount
• Show users with the highest CPU time use: sreport user topuser

tldr.sh

sreport [Options...] [COMMAND...]

sreport is used to generate reports of job usage and cluster utilization for Slurm jobs saved to the Slurm Database, slurmdbd.

-a,  --all_clusters

Use all clusters instead of only the cluster from which the command was executed.

-M,  --cluster

The cluster(s) to generate reports for. Default is local cluster, unless the local cluster is currently part of a federation and in that case generate a report for all clusters in the current federation. If the clusters included in a federation vary through time, use this option to identify the clusters to be included in report. Implies --local.

--federation

Generate a report for the federation if a member of one.

-h,  --help

Print a help message describing the usage of sreport.

--local

Generate a report for the local cluster, even if part of a federation. Overrides --federation.

-n,  --noheader

Don't display header when listing results.

-p,  --parsable

Output will be '|' delimited with a '|' at the end.

-P,  --parsable2

Output will be '|' delimited without a '|' at the end.

-Q,  --quiet

Print no warning or informational messages, only error messages.

-t <time_format>

Specify the output time format. Time format options are case insensitive and may be abbreviated. The default format is Minutes. Supported time format options are listed in the time command section below.

-T,  --tres=<tres_names>

Trackable resource (TRES) to report values for. By default CPU resource use is reported (except for reservation reports. All TRES types reserved by a reservation will be shown by default unless overridden with this option). Multiple TRES names may be separated using a comma separated list for all reports except the job reports, which can only support a single TRES name, or "ALL" for all TRES.

The "Reported" Billing TRES is calculated from the largest Billing TRES of each node multiplied by the time frame. For example, if a node is part of multiple partitions and each has a different TRESBillingWeights defined the Billing TRES for the node will be the highest of the partitions. If TRESBillingWeights is not defined on any partition for a node then the Billing TRES will be equal to the number of CPUs on the node.

TRES node usage is no longer reported in percent format or in Cluster Utilization. Please use TRES CPU instead. The main issue with using node is in most configurations multiple jobs are able to run on the same node. This makes TRES node accounting count the same node multiple times in the same period. In exclusive node configurations, CPU accounting returns the same usage node accounting would.

-v,  --verbose

Print detailed event logging.

-V , --version

Print version information and exit.

<keyword> may be omitted from the execute line and sreport will execute in interactive mode. sreport will process commands as entered until explicitly terminated.

exit

Terminate the execution of sreport. Identical to the quit command.

help

Display a description of sreport options and commands.

nonparsable

Return output to normal after parsable or parsable2 has been set.

parsable

Output will be | delimited with an ending '|'.

parsable2

Output will be | delimited without an ending '|'.

quiet

Print no warning or informational messages, only fatal error messages.

quit

Terminate the execution of sreport. Identical to the exit command.

time <time_format>

Specify the output time format. Time format options are case insensitive and may be abbreviated. The default format is Minutes. Supported time format options include:

SecPer

Seconds/Percentage of Total

MinPer

Minutes/Percentage of Total

HourPer

Hours/Percentage of Total

Seconds

Seconds

Minutes

Minutes

Hours

Hours

Percent

Percentage of Total

verbose

Enable detailed event logging.

version

Display the sreport version number.

!!

Repeat the last command executed.

Valid report types are:

<REPORT> options for each type include:

cluster

AccountUtilizationByUser, UserUtilizationByAccount, UserUtilizationByWckey, Utilization, WCKeyUtilizationByUser

job

SizesByAccount, SizesByAccountAndWcKey, SizesByWckey

reservation

Utilization

user

TopUsage

NOTE: If OverSubscribe is configured to FORCE or YES in your slurm.conf and the system is not configured to use preemption (PreemptMode=OFF) accounting can easily grow to values greater than the actual utilization. It may be common on such systems to get error messages in the slurmdbd log stating: "We have more allocated time than is possible."

cluster AccountUtilizationByUser

This report will display account utilization as it appears on the hierarchical tree. Starting with the specified account or the root account by default this report will list the underlying usage with a sum on each level. Use the 'tree' option to span the tree for better visibility.

NOTE: Idle reservation time will be split evenly among accounts/users given access to it. When a reservation is assigned to whole accounts, the time will be counted in the association for the accounts, not the user associations in the accounts. In this case, the usage of a parent account can be larger than the sum of its children.

cluster UserUtilizationByAccount

This report will display users by account in order of utilization without grouping multiple accounts by user into one, but displaying them on separate lines.

cluster UserUtilizationByWCKey

This report will display users by wckey in order of utilization without grouping multiple wckey by user into one, but displaying them on separate lines.

cluster Utilization

This report will display total usage divided amongst Allocated, Down, Planned Down, Idle, and Planned time for selected clusters.

Allocated

Time that nodes were in use with active jobs or an active reservation. This does not include reservations created with the MAINT or IGNORE_JOBS flags.

Down

Time that nodes were marked as Down or fully Drained, or time that slurmctld was not responding (assuming TrackSlurmctldDown is set in slurmdbd.conf).

Idle

Time that nodes were not Allocated, Down, PlannedDown, or Planned.

OverCommitted

Time of eligible jobs waiting in the queue over the Planned time. Unlike Planned, this has no limit. It is typically useful to determine whether your system is overloaded and by how much.

PlannedDown

Time that nodes were in use by a reservation created with the MAINT flag but not the IGNORE_JOBS flag. Also, time that nodes were in the FUTURE or POWERED_DOWN state.

Planned

Time that nodes were not Allocated, Down or PlannedDown with eligible jobs in the queue that were unable to start due to time or size constraints. The maximum time is limited to the amount of time possible on the system. If this value is not of importance for you then the number can be grouped with idle time.

TresCount

Configured number of TRESs on the reported nodes. See also the TresName field.

Note: Reservations created with the IGNORE_JOBS flag are not tracked in the Cluster Utilization report due to the fact that allowing any current/active jobs to continue to run in the reservation introduces the possibility for them to be accounted for incorrectly. The jobs in these reservations will be tracked as normal rather than being bundled in the reservation time, as they are with reservations that do not have the IGNORE_JOBS flag.
Note: The default view for the "Cluster Utilization" report includes the following fields: Cluster, Allocated, Down, PlannedDown, Idle, Planned, Reported. You can include additional fields like OverCommitted and TresCount fields with the Format option. The TresName will also be included if using the -T, --tres <tres_names> option.

cluster WCKeyUtilizationByUser

This report will display wckey utilization sorted by WCKey name for each user on each cluster.

job SizesByAccount

This report will display the amount of time used for job ranges specified by the 'grouping=' option. Only a single level in the tree is displayed defaulting to the root dir. If you specify other accounts with the 'account=' option sreport will use those accounts as the root account and you will receive the aggregated totals of each listed account plus their sub accounts.

job SizesByAccountAndWckey

This report is very similar to SizesByAccount with the difference being each account is pair with wckeys so the identifier is account:wckey instead of just account so there will most likely be multiple accounts listed depending on the number of wckeys used.

job SizesByWckey

This report will display the amount of time for each wckey for job ranges specified by the 'grouping=' option.

reservation Utilization

This report will display total usage for reservations on the systems. Note: Time requests on this report will not truncate the time the reservation used, only the reservations that ran at any time during the period requested.

user TopUsage

Displays the top users on a cluster, i.e. users with the highest usage. By default users are sorted by CPUTime, but the -T, --tres option will sort users by the first TRES specified.

Use the group option to group accounts together. The default is to have a different line for each user account combination.

Each report type has various options...

Options FOR ALL Report Types

All_Clusters

Use all monitored clusters. Default is local cluster.

Clusters=<OPT>

List of clusters to include in report. Default is local cluster.

End=<OPT>

Period ending for report. Default is 23:59:59 of previous day. Valid time formats are...

HH:MM[:SS] [AM|PM]
MMDD[YY] or MM/DD[/YY] or MM.DD[.YY]
MM/DD[/YY]-HH:MM[:SS]
YYYY-MM-DD[THH:MM[:SS]]
now[{+|-}count[seconds(default)|minutes|hours|days|weeks]]

Format=<OPT>

Comma separated list of fields to display in report.

When using the format option for listing various fields you can put a %NUMBER afterwards to specify how many characters should be printed.

e.g. format=name%30 will print 30 characters of field name right justified. A -30 will print 30 characters left justified.

Start=<OPT>

Period start for report. Default is 00:00:00 of previous day. Valid time formats are...

HH:MM[:SS] [AM|PM]
MMDD[YY] or MM/DD[/YY] or MM.DD[.YY]
MM/DD[/YY]-HH:MM[:SS]
YYYY-MM-DD[THH:MM[:SS]]
now[{+|-}count[seconds(default)|minutes|hours|days|weeks]]

Options SPECIFICALLY FOR CLUSTER REPORTS

Accounts=<OPT>

When used with the UserUtilizationByAccount, or AccountUtilizationByUser, List of accounts to include in report. Default is all.

Tree

When used with the AccountUtilizationByUser report will span the accounts as they are in the hierarchy.

Users=<OPT>

When used with any report other than Utilization, List of users to include in report. Default is all.

Wckeys=<OPT>

When used with the UserUtilizationByWckey or WCKeyUtilizationByUser, List of wckeys to include in report. Default is all.

Options SPECIFICALLY FOR JOB REPORTS

Accounts=<OPT>

List of accounts to use for the report. Default is all which will show only one line corresponding to the totals of all accounts in the hierarchy. This explanation does not apply when ran with the FlatView or AcctAsParent options.

AcctAsParent

When used with the SizesbyAccount(*) will take specified accounts as parents and the next layer of accounts under those specified will be displayed. Default is root if no specific Accounts are requested. When FlatView is used, this option is ignored.

FlatView

When used with the SizesbyAccount(*) will not group accounts in a hierarchical level, but print each account where jobs ran on a separate line without any hierarchy.

GID=<OPT>

List of group ids to include in report. Default is all.

Grouping=<OPT>

Comma separated list of size groupings. (e.g. 50,100,150 would group job cpu count 1-49, 50-99, 100-149, > 150). grouping=individual will result in a single column for each job size found.

Jobs=<OPT>

List of jobs/steps to include in report. Default is all.

Nodes=<OPT>

Only show jobs that ran on these nodes. Default is all.

Partitions=<OPT>

List of partitions jobs ran on to include in report. Default is all.

PrintJobCount

When used with the Sizes report will print number of jobs ran instead of time used.

Users=<OPT>

List of users jobs to include in report. Default is all.

Wckeys=<OPT>

List of wckeys to use for the report. Default is all. The SizesbyWckey report all users summed together. If you want only certain users specify them with the Users= option.

Options SPECIFICALLY FOR RESERVATION REPORTS

Names=<OPT>

List of reservations to use for the report. Default is all.

Nodes=<OPT>

Only show reservations that used these nodes. Default is all.

Options SPECIFICALLY FOR USER REPORTS

Accounts=<OPT>

List of accounts to use for the report. Default is all.

Group

Group all accounts together for each user. Default is a separate entry for each user and account reference.

TopCount=<OPT>

Used in the TopUsage report. Change the number of users displayed. Default is 10.

Users=<OPT>

List of users jobs to include in report. Default is all.

FORMAT Options FOR CLUSTER REPORTS

AccountUtilizationByUser

Accounts, Cluster, Login, Proper, TresCount, Used

UserUtilizationByAccount

Accounts, Cluster, Login, Proper, TresCount, Used

UserUtilizationByWckey

Cluster, Login, Proper, TresCount, Used, Wckey

Utilization

Allocated, Cluster, Down, Idle, OverCommitted, PlannedDown, Reported, Planned, TresCount, TresName

WCKeyUtilizationByUser

Cluster, Login, Proper, TresCount, Used, Wckey

FORMAT Options FOR JOB REPORTS

SizesByAccount

Account, Cluster

SizesByAccountAndWckey

Account, Cluster

SizesByWckey

Wckey, Cluster

FORMAT Options FOR RESERVATION REPORTS

Utilization

Allocated, Associations, Cluster, End, Flags, Idle, Name, Nodes, ReservationId, Start, TotalTime, TresCount, TresName, TresTime

FORMAT Options FOR USER REPORTS

TopUsage

Account, Cluster, Login, Proper, Used

All commands and options are case-insensitive.

Executing sreport sends a remote procedure call to slurmdbd. If enough calls from sreport or other Slurm client commands that send remote procedure calls to the slurmdbd daemon come in at once, it can result in a degradation of performance of the slurmdbd daemon, possibly resulting in a denial of service.

Do not run sreport or other Slurm client commands that send remote procedure calls to slurmdbd from loops in shell scripts or other programs. Ensure that programs limit calls to sreport to the minimum necessary for the information you are trying to gather.

Some sreport options may be set via environment variables. These environment variables, along with their corresponding options, are listed below. (Note: Command line options will always override these settings.)

SREPORT_CLUSTER

Same as -M, --cluster

SREPORT_FEDERATION

Same as --federation

SREPORT_LOCAL

Same as --local

SREPORT_TRES

Same as -t, --tres

SLURM_CONF

The location of the Slurm configuration file.

Report number of jobs per account according to different job size bins:

$ sreport job sizesbyaccount

Report cluster utilization:

$ sreport cluster utilization

Report top usage:

$ sreport user top

Report number of jobs by user gore1 within the environ account:

$ sreport job sizesbyaccount All_Clusters users=gore1 account=environ PrintJobCount

Report cluster account utilization with the specified fields during the specified 24 hour day of February 23, 2009, by user gore1:

$ sreport cluster AccountUtilizationByUser cluster=zeus user=gore1 start=2/23/09 end=2/24/09 format=Accounts,Cluster,TresCount,Login,Proper,Used

Report cluster account utilization by user in the LC account on cluster zeus:

$ sreport cluster AccountUtilizationByUser cluster=zeus accounts=lc start=2/23/09 end=2/24/09

Report top usage in percent of the lc account during the specified week:

$ sreport user topusage start=2/16/09 end=2/23/09 -t percent account=lc

Copyright (C) 2009-2010 Lawrence Livermore National Security. Produced at Lawrence Livermore National Laboratory (cf, DISCLAIMER).
Copyright (C) 2010-2022 SchedMD LLC.

This file is part of Slurm, a resource management program. For details, see <https://slurm.schedmd.com/>.

Slurm is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.

Slurm is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

sacct(1), slurmdbd(8)

slurm(1).