MXG Software separates a true JCL error (i.e. prevents EXECUTION)
from an "execution" JCL error, because the latter jobs did initiate
and are thus valid to be included in job service level reporting,
whereas the inclusion of JCL errors in those reports is wrong.

(However, if they were incorrectly included in your job reporting,
you could easily improve the daily statistics by submitting
a few hundred JCL error jobs!)

The full suite of "ABEND" causes decoded from the SMF 26 and 30 
records for each job are these possible values for MXG variables 
ABEND and CONDCODE:


Variable Type Length  Format       Label

ABEND     CHAR   6              COMPLETION*ABEND*INDICATOR
  Step completion indicator.  A description of how or why this step
  ended. For non-normal completions, see associated variable CONDCODE
  for amplification of the termination reason.

  value     meaning

  blank     normal completion.
  CANCEX    step was canceled by an SMF exit; variable CONDCODE contains
              a bit indicating the exit that cancelled the step.
  COND=J    step was not executed because of COND= on the step EXEC.
  FLUSH     step was flushed (because a previous step had failed).
  NOTCTL    step completed with a "post execution" error, which includes
              either NOT CATALOGED 2 or NOT CATALOGED 7 condition, only
              if your installation specified cancellation in the PARMLIB
              member controlling allocation. See APAR OY38977.
  OMVSEX    step record written because address changed state to execute
            OMVS (Open Edition/MVS, IBM's MVS UNIX).
  OTHER     step abended for unknown reason. Variable CONDCODE  contains
              the value in register 15 at time of the abend.  "OTHER" is
              set when the step ABEND bit is on but the numerical value
              of SMF30SCC (the step completion code) does not describe
              the reason for the step ABEND bit in SMF30STI.
  RESTAR     step was restarted; subsequent restart has program name of
              IEFRSTRT.
  RETURN    step completed with nonzero condition code. CONDCODE has the
              return code value.
  SYSTEM    step completed with system abnormal end (abend).  CONDCODE
              has the abend code.
  USER      Step completed with an abend issued by a user program,
              rather than by the operating system itself. Any programmer
              can choose to issue a USER ABEND rather than a return code
              (for example, the SAS System issues a USER ABEND 999 when
              ERRORABEND is specified and an error occurs).  CONDCODE
              contains the abend code.

  Note: The preceding values for ABEND are those set in processing the
        type 30 record.  In the BUILDPDB process, the PDB.JOBS dataset
        carries forward the variable ABEND, and sets additional values
        for the variable ABEND.  Although these values will not be seen
        in the TYPE30 datasets, they are listed here (and also in member
        ADOCPDB) for clarification:

  CRSH      Job was active when system "Crashed" (i.e., the JSTRTIME in
            the TYPE26J2 record was missing or earlier than the JENDTIME
            indicating the job had started but never returned from MVS).
  JCL       Job had a JCL error prior to execution (i.e., the job never
            began execution, either due to a JCL error detected in the
            Reader-Interpreter or Converter, or the job was cancelled by
            the operator or user, in which case CANCELED='Y'.)  This
            value of ABEND is set so that these jobs can be excluded
            from job scheduling statistics (since these jobs have nearly
            instantaneous input queue time, including them in your batch
            statistics is wrong - you could make the daily stats look
            real good by submitting thousands of JCL errors!).
            Note that jobs with what I call an "execution JCL error",
            like Data Set Not Found, that actually occur after the job
            has started are NOT set with ABEND='JCL' (even though the
            user gets a NOTIFY message that the job had a JCL error)
            because those jobs were in fact scheduled and executed by
            the data center and thus should be included in statistics.
  PREEXE    JOBS-only: Step had a pre-execution ABEND. 



CONDCODE  NUM    5 HEX4.0       COMPLETION*CONDITION*CODE
  Condition code. This variable amplifies the meaning of ABEND.  It is
  internally a binary number but is formatted to display as hex so that
  testing for a specific condition uses this syntax:

       IF ABEND= 'SYSTEM' AND CONDCODE=0322X THEN ...

  For some values of ABEND, CONDCODE has these meanings:
    ABEND       CONDCODE
    value       value

    CANCEXIT      4 = Cancelled by IEFUJI (job initiation) SMF exit
    CANCEXIT      2 = Cancelled by IEFUSI (step initiation) SMF exit
    CANCEXIT      1 = Cancelled by IEFACTRT (step termination) SMF exit
    OTHER        completion code value
    RETURN       return code value
    SYSTEM       system ABEND code
    USER         user ABEND code.
  In z/OS 1.13, the JOB card can be used to indicate which condition
  code you want in your type30 subtype 5 (TYPE30_5) SMF record, last
  highest or a named step.

Barry Merrill



-----Original Message-----
From: IBM Mainframe Discussion List [mailto:[email protected]] On Behalf 
Of Shmuel Metz (Seymour J.)
Sent: Monday, October 14, 2013 8:35 AM
To: [email protected]
Subject: Re: SYSIN Symbol Resolution vs. LRECL

In <[email protected]>, on
10/13/2013
   at 11:00 AM, Paul Gilmartin <[email protected]> said:

>"DATA SET NOT FOUND" is a JCL ERROR (Read the IEF453I message text),

I know what the message text says. Now consider a job with two steps; the first 
executes a program that deletes a data set, the second has a DD statement for 
the deleted dataset with DISP=OLD. The job fails in the second step only 
because of what the first step did. Is the error in the JCL or in the input to 
the first program?
 
-- 
     Shmuel (Seymour J.) Metz, SysProg and JOAT
     ISO position; see <http://patriot.net/~shmuel/resume/brief.html>
We don't care. We don't have to care, we're Congress.
(S877: The Shut up and Eat Your spam act of 2003)

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions, send email to 
[email protected] with the message: INFO IBM-MAIN

----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [email protected] with the message: INFO IBM-MAIN

Reply via email to