Re: Need to record the timing of each CALL step in a batch job...

[CRPence <CRPbottle@xxxxxxxxx>]

On 12/8/10 1:32 PM, Charles Wilt wrote:
> I have a request to start tracking the time each step takes in a
> batch process...
>
> A manual process might be, to look through the job log for
>
> *NONE      Command    12/08/10  02:29:44.407088
> QCLCLCPR     QSYS        04CB     FS001C      HLTHPRDPGM  0242
>                 Message . . . . :    34100 - CALL PGM(PR002R)
>                      /* The CALL command contains parameters */
>
> *NONE      Command    12/08/10  02:29:57.622312
> QCLCLCPR     QSYS        04CB     FS001C      HLTHPRDPGM  0251
>                 Message . . . . :    34900 - CALL  PGM(PR051R)
>                      /* The CALL command contains parameters */
>
> Obviously, a manual process would not be my preferred method, given
> that this needs to be done for multiple jobs and is not a one-time
> thing. Additionally, I need the specifics of the parameters used
> during the call.
>
> My thoughts
> 1) Use command exit point(s) process/log the CALLs, but it appears
> that CALL is one command for which the exit point is not invoked.
> 2) Replace CALL command with custom command whose CPP does the
> logging needed
> 3) DSPJOBLOG OUTPUT(*FILE) or maybe OUTPUT(*APIDFN) at the end of the
> process, but *FILE at least doesn't provide the parameters used
> during the call
> 4) List Job Log Messages (QMHLJOBL) API or Open List of Job Log
> Messages (QGYOLJBL) API, guessing that neither of these will have the
> parameter info I need either...
>
> So it appear that my only option is option 2....
>
> Does anybody have any thoughts, comments or alternatives?

  When I have had need to do something like the described for a 
specific process [e.g. a test-case or other batch script or CLP], I 
typically would convert the CL or CLP to REXX to have a procedure 
DoRequest which is then invoked by "DoRequest CALL PGM(...", having 
replaced each "CALL PGM(...".  The DoRequest procedure first does a SAY 
of the request, then the request itself, and finally a summary SAY; the 
former adding something like TIME() as a prefix.

  FWiW, seems to me that "5) Application logging" is a better option if 
the function might best, not be limited to just the one script\program:

  One approach is to change the calling program(s) to invoke some 
logging prior to [and probably also after] each call.  Another approach 
has each called program inquire [by review either of a passed parameter 
or of some logging indication] to determine if a request for logging is 
currently active, then log entry and optionally exit.

Regards, Chuck