NAME
     accounting - Grid Engine accounting file format

DESCRIPTION
     An accounting record is written to the Grid Engine  account-
     ing  file  for each job having finished. The accounting file
     is processed by qacct(1) to derive accounting statistics.

FORMAT
     Each job is represented by a line in  the  accounting  file.
     Empty  lines  and  lines which contain one character or less
     are ignored.  Accounting record  entries  are  separated  by
     colon  (':')  signs.  The  entries  denote in their order of
     appearance:

  qname
     Name of the cluster queue in which the job has run.

  hostname
     Name of the execution host.

  group
     The effective group id of the job owner when  executing  the
     job.

  owner
     Owner of the Grid Engine job.

  job_name
     Job name.

  job_number
     Job identifier - job number.

  account
     An account string as specified by the qsub(1)  or  qalter(1)
     -A option.

     priority
          Priority value assigned to the job corresponding to the
          priority  parameter  in  the  queue  configuration (see
          queue_conf(5)).

  submission_time
     Submission time (GMT unix time stamp).  For slave  tasks  of
     tightly  integrated  parallel  jobs,  the submission_time is
     always set to 0.

  start_time
     Start time (GMT unix time stamp).


  end_time
     End time (GMT unix time stamp).

  failed
     Indicates the problem which occurred in case a job could not
     be  started on the execution host (e.g. because the owner of
     the job did not have a valid account on  that  machine).  If
     Grid  Engine  tries  to start a job multiple times, this may
     lead to multiple entries in the accounting file  correspond-
     ing to the same job ID.

  exit_status
     Exit status of the  job  script  (or  Grid  Engine  specific
     status in case of certain error conditions).

  ru_wallclock
     Difference between end_time and start_time (see above).

     The remainder of the accounting entries follows the contents
     of  the  standard  UNIX  rusage  structure  as  described in
     getrusage(2).  Depending on the operating system  where  the
     job was executed some of the fields may be 0.  The following
     entries are provided:

          ru_utime
          ru_stime
          ru_maxrss
          ru_ixrss
          ru_ismrss
          ru_idrss
          ru_isrss
          ru_minflt
          ru_majflt
          ru_nswap
          ru_inblock
          ru_oublock
          ru_msgsnd
          ru_msgrcv
          ru_nsignals
          ru_nvcsw
          ru_nivcsw

  project
     The project which was assigned to the job.

  department
     The department which was assigned to the job.

  granted_pe
     The parallel environment which was selected for that job.


  slots
     The number of slots which were dispatched to the job by  the
     scheduler.

  task_number
     Array job task index number.

  cpu
     The cpu time usage in seconds.

  mem
     The integral memory usage in Gbytes cpu seconds.

  io
     The amount of data transferred in input/output operations.

  category
     A string specifying the job category.

  iow
     The io wait time in seconds.

  pe_taskid
     If this identifier is set the task was part  of  a  parallel
     job  and  was  passed  to  Grid Engine via the qrsh -inherit
     interface.

  maxvmem
     The maximum vmem size in bytes.

SEE ALSO
     sge_intro(1), qacct(1),  qalter(1),  qsub(1),  getrusage(2),
     queue_conf(5).

COPYRIGHT
     See sge_intro(1) for a full statement of rights and  permis-
     sions.
















Man(1) output converted with man2html