The NSCHED process terminates without error shortly after startup

Products

• System Watchdog for OpenVMS

Issue/Introduction

Shortly after executing the CA Job Management Manager for OpenVMS startup procedure, UJM$MANAGER$STARTUP.COM, you find the NSCHED server process terminates and there are no errors in its log file.

Debug tracing is then enabled by defining the logical name NSCHED$DEBUG to a value of "TRUE" and Job Management restarted by executing  UJM$MANAGER$STARTUP.COM. Shortly thereafter, the NSCHED process terminates again but with substantially more information in its log file.  Analysis of the log file shows that a consistency check is being performed on the job scheduling database, which is normal, but again there is no error message displayed. The messages generated by the consistency check look like the example below but with job numbers from your database:

11-MAR-2017 15:24:02.27     queued nostart ast for job  4309  11-MAR-2017 17:30:00.00       
11-MAR-2017 15:24:02.27     03:24 PM  processing record #  4309  status= S   request=  
11-MAR-2017 15:24:02.27      Now=11-MAR-2017 15:24:02.27   job_sched_time=11-MAR-2017 15:30:00.00
11-MAR-2017 15:24:02.27     job  4309  is scheduled for the future
11-MAR-2017 15:24:02.27     updated    record #  4309  status= S    request= 

The final information contained in the log file is standard OpenVMS resource accounting information similar to the example below:

  SYSTEM       job terminated at 11-MAR-2017 15:24:03.24

  Accounting information:
  Buffered I/O count:                327      Peak working set size:      14848
  Direct I/O count:                 3507      Peak virtual size:         244096
  Page faults:                      1320      Mounted volumes:                0
  Charged CPU time:        0 00:00:00.82      Elapsed time:       0 00:00:05.15

 

Environment

OpenVMS Alpha V8.* and OpenVMS I64 V8.*

Cause

The scheduling database consistency check is a very resource intensive operation, and when that database contains several thousand jobs the default values for some resource quotas are insufficient. In the two known situations where this problem has occurred, the default AST limit quota of 400 was not sufficient.

Resolution

Increase both the AST limit and buffered IO byte count quota by creating or editing the file named NSCHED$:UJM$MANAGER$STARTUP_SYSCLUSTER.COM and adding the following two DCL symbol definitions:

$  NSCHED_AST_LIMIT     ==     5000      
$  NSCHED_BUFFER_LIMIT  ==  500000

Both of the above values can be increased if desired, but we do not recommend using lower values if the problem has occurred on your system.

 

Additional Information

Enabling the debug trace for the NSCHED process can be accomplished by executing UJM$MANAGER$STARTUP.COM with the character string DEBUG specified as a parameter as shown in the example below:

$@SYS$STARTUP:UJM$MANAGER$STARTUP  DEBUG