Monitoring Executing Jobs

This forum provides the support of Dezhi Mainframe systems. Please post your questions about logon, usage of our mainframe environment.

Moderators: sysprog, prino, sfan, steve-myers, Tim001

Monitoring Executing Jobs

Postby steve-myers » Tue 25 Dec 2012, 01:42

The admins get very unhappy when they discover jobs that are running too long. In the Fandezhi environment this means jobs using thousands of seconds CPU time. The admins encourage users to
  • Restrict the CPU time a job can use, or
  • Monitor your job while it is executing and cancel the job if it starts behaving unexpectedly.
Either bullet can be easily accomplished.

Restrict CPU time

You restrict the CPU time your job can use by specifying the TIME=(mm,ss) parameter on the JOB JCL statement or the // EXEC statement of the step you expect to run long. When you create a userid, Fandezhi creates a userid.JCL.CNTL dataset with one member: JOBCARD. This member looks like this -
Code: Select all
000001 //ADUMMYJ  JOB (12345678),ADUMMY,MSGCLASS=H,                     
000002 //        MSGLEVEL=(1,1),CLASS=A,NOTIFY=&SYSUID                   
000003 //***************************************************************
000004 //***************JCL CREATED BY A DUMB USER  *********************
000005 //***************************************************************

For better or worse, Fandezhi does not have any default restrictions, in part because job execution on Fandezhi requires much, much more CPU time than you may be used to running jobs on a real mainframe. Still, very few programs require more than a few seconds CPU time. It might be a good idea to update the default like this.
Code: Select all
000001 //ADUMMYJ  JOB (12345678),ADUMMY,MSGCLASS=H,TIME=5,                     
000002 //        MSGLEVEL=(1,1),CLASS=A,NOTIFY=&SYSUID                   
000003 //***************************************************************
000004 //***************JCL CREATED BY A DUMB USER  *********************
000005 //***************************************************************

This restricts the job to a maximum of 5 minutes, or 300 seconds, CPU time, which, hopefully, will not upset the admins.

Monitor your job while it is executing

It's really easy to monitor your job while it is executing. Just start SDSF, establish a job prefix if you haven't done this already, and enter DA on the command line and press your Enter key. SDSF will display something like this.
Code: Select all
  Display  Filter  View  Print  Options  Help                                 
------------------------------------------------------------------------------
SDSF DA SYS1  ADCD     PAG    0 SIO     0 CPU  14/***  LINE 1-2 (2)           
COMMAND INPUT ===>                                            SCROLL ===> PAGE
NP   JOBNAME  StepName ProcStep JobID    Owner    C Pos DP Real Paging    SIO
     ADUMMY   SYSUSER  TCPA0335 TSU97463 ADUMMY     IN  F7 1291   0.00   0.00
Press PF11 once, and SDSF will display something like this.
Code: Select all
SDSF DA SYS1  ADCD     PAG    0 SIO     0 CPU  68/***  LINE 1-3 (3)           
COMMAND INPUT ===>                                            SCROLL ===> PAGE
NP   JOBNAME   CPU% ASID ASIDX  EXCP-Cnt   CPU-Time SR Status SysName  SPag SCP
     ADUMMYAJ 53.50   39  0027         6      18.20           ADCD        0   
     ADUMMY   26.34  110  006E   132,944     313.77           ADCD        0

Don't worry about the CPU% column; even very I/O bound programs will show a very high amount on Fandezhi. Watch the CPU-time and EXCP-Cnt columns. If the CPU time goes up and the EXCP count stays the same or goes up slowly, something may be wrong. You can display the JES2 datasets by entering ? in the NP column and pressing your Enter key. You can then select a data set in the usual way.
steve-myers
 
Posts: 456
Joined: Tue 04 May 2010, 15:43

Return to Dezhi systems: Mainframe

Who is online

Users browsing this forum: No registered users and 2 guests

cron