Scheduler jobs tracking problem can occur when JES is slow
Article ID: 25038
Updated On:
Products
Issue/Introduction
Scheduler r11 PTF RO05022 addresses jobs tracking problems that may result when JESx performs slowly. A formula is available to gauge the performance of JESx using the Scheduler statistics.
Cause
Scheduler uses the IBM ENDREQ macro to complete submission and retrieve the JESx job number for batch jobs. When JESx is performing slowly, the ENDREQ service may take a long time to complete, and jobs can start execution before Scheduler is "finished" submitting the job. This results in tracking issues because Scheduler receives a JOBINIT event from CAIENF before Scheduler finishes submitting that job.
Resolution
1. Run utility program CAJUTIL0 to obtain the statistics for SUB ENDREQ TIME and SUB JOBS SUBBED.
//STEP1 EXEC CAJUTIL0 //SYSIN DD * REPORT STATS ID='SUB ENDREQ TIME',I=1440,FD=mmddyyyy,TD=mmddyyyy REPORT STATS ID='SUB JOBS SUBBED',I=1440,FD=mmddyyyy,TD=mmddyyyy //
Use an Interval of 1440 to summarize the statistics for the entire day. You may include a range of dates or just a single date for the From Day and To Day fields. Sites with multiple LPARs should also include SYSID= on the command pointing to the master copy of Scheduler.
The output will look something like this:
REPORT STATS ID='SUB ENDREQ TIME',I=1440,FD=03132009,TD=03132009
Statistic Name Description
---------------- ----------------------------------------------
SUB ENDREQ TIME Time needed to process ENDREQ at end of job
submission.
STATISTIC NAME=SUB ENDREQ TIME FOR DATE=03/13/2009 SYSID=SYSA
0000-2359 2866313 11740.418048 SECONDS
REPORT STATS ID='SUB JOBS SUBBED',I=1440,FD=03132009,TD=03132009
Statistic Name Description
---------------- ----------------------------------------------
SUB JOBS SUBBED Number of normal batch jobs submitted.
STATISTIC NAME=SUB JOBS SUBBED FOR DATE=03/13/2009 SYSID=SYSA
0000-2359 44960
2. The formula described in PTF RO05022 is as follows:
'SUB ENDREQ TIME' divided by 'SUB JOBS SUBBED' times .004096 = seconds per job required by ENDREQ
In this example, we see:
2866313 / 44960 x .004096 = .26 seconds
Each job required an average of .26 seconds to finish being submitted and for JESx to return a job number to Scheduler.
A quarter second is a very long time, making tracking problems (without PTF RO05022) more likely.
Here's another example:
REPORT STATS ID='SUB ENDREQ TIME',I=1440,FD=03192009,TD=03192009
Statistic Name Description
---------------- ----------------------------------------------
SUB ENDREQ TIME Time needed to process ENDREQ at end of job
submission.
STATISTIC NAME=SUB ENDREQ TIME FOR DATE=03/19/2009 SYSID=SYSA
0000-2359 358965 1470.320640 SECONDS
REPORT STATS ID='SUB JOBS SUBBED',I=1440,FD=03192009,TD=03192009
Statistic Name Description
---------------- ----------------------------------------------
SUB JOBS SUBBED Number of normal batch jobs submitted.
STATISTIC NAME=SUB JOBS SUBBED FOR DATE=03/19/2009 SYSID=SYSA
0000-2359 17515
In this case:
358965 / 17515 x .004096 = .08 seconds, or less than a tenth of a second.
JESx performance can be influenced by a number of factors from data set placement to the amount of work on the system to WLM settings.
Scheduler statistics can be a useful tool in gauging your system performance.
Feedback
