qacct

qacct is a scheduler command used to display accounting data for all jobs and job steps in the SGE job accounting file / SGE accounting database.

Documentation

Documentation is available on the system using the command

$ man qacct

Usage

The qacct command can be used to display status information about a user’s historical jobs.

Running the qacct command alone will provide a summary of used resources from the current month for the user running the command.

The command can be used as follows with the job’s ID to get job specific info:

$ qacct -j job-id

Or to view information about all of a specific user’s jobs:

$ qacct -j -u $USER

Important metrics listed for your jobs are detailed in the table below:

qacct property variable names and descriptions

Variable	Description
qname	The queue the job ran in.
hostname	The hostname of the node used as the master.
owner	The username of the person who ran the job.
jobname	The name of the job.
jobnumber	The Job ID number.
taskid	The taskid of the job if a task array job.
granted_pe	The parallel environment the job ran in.
slots	The number of cores used for the job.
failed	A binary value indicating whether the program or script ran in the job exited normally or in and error state.
exit_status	The exit status or code given by the submission script when it finished executing.
ru_wallclock	The wallclock time of the job (i.e. if you looked at a clock on the wall and noted how long the job took.)
maxvmem	The maximum memory used during the job.
category	Contains the submission arguments used for qsub.
io	The amount of read / write to or from storage areas.
iow	The amount of time the job was waiting for io, i.e.waiting for a storage to be read from or wrtten to.

By default the qacct command will only bring up summary info about the user’s jobs from the current accounting file (which rotates monthly). By using the -f AcctFilePath flag and the accounting file path of interest, previous months usage can be quantified e.g. :

qacct -u $USER -f $SGE_ROOT/default/common/accounting-archive/accounting-20210101

By using bash xargs and find you can also parse multiple files to list jobs as follows:

find $SGE_ROOT/default/common/accounting-archive/ -mtime -90 -type f | xargs -t -n1 qacct -j -u $USER -f

This command is instructing the linux find executable to look for files younger than 90 days and then xargs passes the accounting files it finds to the qacct command (in bold.)

Like previous qacct commands the -j flag can once again take a job ID to pull up a specific job’s details.