[Chapter 13] 13.3 Tips on Using DBMS

13.3 Tips on Using DBMS_JOB

This section discusses several useful tips for using DBMS_JOB.

13.3.1 Job Intervals and Date Arithmetic

Job execution intervals are determined by the date expression set by the interval parameter. Getting jobs to run at the desired times can be one of the more confusing aspects of using DBMS_JOB and the job queue. One key to setting the interval correctly is determining which of the following applies to the job:

Each execution of the job should follow the last by a specific time interval.
The job should execute on specific dates and times.

Jobs of type 1 usually have relatively simple date arithmetic expressions of the type SYSDATE+N, where N represents the time interval expressed in days. The following table provides examples of these types of intervals.

Action	Interval Value
Execute daily	'SYSDATE + 1'
Execute hourly	'SYSDATE + 1/24'
Execute every 10 minutes	'SYSDATE + 10/1440'
Execute every 30 seconds	'SYSDATE + 30/86400'
Execute every 7 days	'SYSDATE + 7'
Do not re-execute and remove job	NULL

Remember that job intervals expressed as shown in the previous table do not guarantee that the next execution will happen at a specific day or time, only that the spacing between executions will be at least that specified. For instance, if a job is first executed at 12:00 p.m. with an interval of SYSDATE + 1, it will be scheduled to execute the next day at 12:00 p.m. However, if a user executes the job manually at 4:00 p.m. using DBMS_JOB.RUN, then it will be rescheduled for execution at 4:00 p.m. the next day. Another possibility is that the database is down or the job queue so busy that the job cannot be executed exactly at the time scheduled. In this case, the job will run as soon as it can, but the execution time will have migrated away from the original submission time due to the later execution. This "drift" in next execution times is characteristic of jobs with simple interval expressions.

Jobs with type 2 execution requirements involve more complex interval date expressions, as seen in the following table.

Action	Interval Value
Every day at 12:00 midnight	'TRUNC(SYSDATE + 1)'
Every day at 8:00 a.m.	'TRUNC(SYSDATE + 1) + 8/24'
Every Tuesday at 12:00 noon	'NEXT_DAY(TRUNC(SYSDATE ), "TUESDAY" ) + 12/24'
First day of the month at midnight	'TRUNC(LAST_DAY(SYSDATE ) + 1)'
Last day of the quarter at 11:00 p.m.	'TRUNC(ADD_MONTHS(SYSDATE + 2/24, 3 ), 'Q' ) - 1/24'
Every Monday, Wednesday, and Friday at 9:00 a.m.	NEXT_DAY(SYSDATE, "FRIDAY") ) ) + 9/24'

Specifying intervals like these can get tricky, so be sure that your date arithmetic expression is correct.

I had hoped that another option for evaluating complex job execution intervals would be to write PL/SQL functions with DATE return values that perform the interval calculations. However, my experiments in this area showed that job intervals that call date functions can be successfully submitted to the job queue but may not be properly executed. The SNP processes appeared to have difficulty properly updating the catalog, and jobs became locked in an endless cycle of execution. Perhaps this limitation will be corrected in future releases.

13.3.2 Viewing Job Information in the Data Dictionary

Information about jobs in the job queue is available through several data dictionary views (see Table 13.2 ) created by the catproc.sql script.

Table 13.2: Data Dictionary Views for DBMS_JOB
View Name	Description
DBA_JOBS	All jobs defined to the job queue in this database
DBA_JOBS_RUNNING	All jobs in the database which are currently executing
USER_JOBS	Jobs in the database owned by the current user

Table 13.3 summarizes the various columns in the DBA_JOBS and USER_JOBS views.

Table 13.3: Columns in DBA_JOBS and USER_JOBS Views
Column	Datatype	Description
JOB	NUMBER	Unique identifier of the job
LOG_USER	VARCHAR2(30)	User who submitted the job
PRIV_USER	VARCHAR2(30)	User whose privileges apply to the job
SCHEMA_USER	VARCHAR2(30)	User schema to parse the job under
LAST_DATE	DATE	Last successful execution date
LAST_SEC	VARCHAR2(8)	Hour, minute, and second portion of last_date formatted as HH24:MI:SS
THIS_DATE	DATE	Date current execution began, or NULL if not executing
THIS_SEC	VARCHAR2(8)	Hour, minute, and second portion of this_date formatted as HH24:MI:SS
NEXT_DATE	DATE	Date of next scheduled execution
NEXT_SEC	VARCHAR2(8)	Hour, minute, and second portion of next_date formatted as HH24:MI:SS
TOTAL_TIME	NUMBER	Total elapsed time in seconds for all executions of this job
BROKEN	VARCHAR2(1)	Flag value Y indicates job broken, will not run
INTERVAL	VARCHAR2(200)	Date function used to compute next_date
FAILURES	NUMBER	Number of consecutive unsuccessful executions
WHAT	VARCHAR2(2000)	PL/SQL block executed as the job
CURRENT_SESSION_LABEL	RAW MLSLABEL	Trusted Oracle session label for the job
CLEARANCE_HI	RAW MLSLABEL	Trusted Oracle high clearance for the job
CLEARANCE_LO	RAW MLSLABEL	Trusted Oracle low clearance for the job
NLS_ENV	VARCHAR2(2000)	execution
MISC_ENV	RAW(32)	Other session parameters for job execution

Table 13.4 shows the columns in the DBA_JOBS_RUNNING view.

Table 13.4: Columns in DBA_JOBS_RUNNING View
Column	Datatype	Description
SID	NUMBER	Session ID currently executing the job
JOB	NUMBER	Unique identifier of the job
FAILURES	NUMBER	Number of consecutive unsuccessful executions
LAST_DATE	DATE	Last successful execution date
LAST_SEC	VARCHAR2(8)	Hour, minute, and second portion of last_date formatted as HH24:MI:SS
THIS_DATE	DATE	Date current execution began
THIS_SEC	VARCHAR2(8)	Hour, minute, and second portion of this_date formatted as HH24:MI:SS

The number and size of the columns in DBA_JOBS and USER_JOBS can make them awkward to query interactively. Several examples of useful scripts to run against the job queue dictionary views follow. One thing I usually do is to set my session NLS_DATE_FORMAT to display the full date and time; in this way, I avoid selecting the date and time portions separately. Note that the date columns in these views contain full date values down to the second; the formatted timestamp columns (LAST_SEC, THIS_SEC, NEXT_SEC) are actually derived from them in the views.

This script shows which jobs are currently executing, who owns them, and when they began:

/* Filename on companion disk:

 job2.sql */*
col job_definition format a30 word_wrap
col username format a15

ALTER SESSION SET NLS_DATE_FORMAT='YYYY:MM:DD:HH24:MI:SS';

SELECT  jr.job       job_id
       ,username     username
       ,jr.this_date start_date
       ,what         job_definition
  FROM 
        dba_jobs_running   jr
       ,dba_jobs           j
       ,v$session          s
 WHERE
        s.sid  = jr.sid
   AND  jr.job = j.job
 ORDER BY jr.this_date;

The following script shows failing or broken jobs (i.e., jobs that may need attention):

/* Filename on companion disk: job2.sql */*
col job_owner format a15
col job_definition format a30 word_wrap

SELECT  job          
       ,log_user     job_owner
       ,failures
       ,broken
       ,what         job_definition
  FROM
        dba_jobs
 WHERE
        broken = 'Y' OR NVL(failures,0) > 0 ;

The next script shows jobs queued up to be executed in order of next execution date. Jobs with negative values in the mins_to_exec column indicate that the job queue is not keeping up with its workload and may need extra job queue processes initiated. The script excludes currently executing jobs because next_date will not be updated until the current execution completes.

/* Filename on companion disk: 

job2.sql */*
col job_definition format a30 word_wrap
col username format a15

ALTER SESSION SET NLS_DATE_FORMAT='YYYY:MM:DD:HH24:MI:SS';

SELECT  job
       ,username
       ,next_date
       ,ROUND((next_date - SYSDATE)*24*60)  mins_to_exec
       ,what                         job_definition
  FROM
        dba_jobs
 WHERE
      broken != 'Y' 
  AND job NOT IN 
         (SELECT job 
            FROM dba_jobs_running)
 ORDER BY next_date ASC;

Here is sample output from the preceding script on a system that has a very busy job queue. Job number 10 will be run next but is already 21 minutes late for execution.

    JOB NEXT_DATE           MINS_TO_EXEC JOB_DEFINITION
----- ------------------- ------------ ------------------------------
   10 1997:11:25:17:04:10          -21 load3.loadx.loop_and_execute(5
                                       ,30,'begin
                                       loadx.table_scanner(5,5);end;'
                                       );

    5 1997:11:25:17:25:21            0 load3.loadx.loop_and_execute(1
                                       0,90,'begin
                                       loadx.cpu_hog(20,20,20);end;')
                                       ;

   12 1997:11:25:17:29:08            4 load2.loadx.loop_and_execute(2
                                       0,60,'begin
                                       loadx.grow_table(''LOAD2'',''T
                                       ABLE2'',500,500);end;');

13.3.3 DBMS_IJOB: Managing Other Users' Jobs

One of the most frustrating aspects of the DBMS_JOB package for DBAs is that its procedures can be executed only against jobs owned by the current user. Even the SYS user cannot remove or set the broken flag for other user's jobs. Thus, job queue environments with multiple job owners can become problematic to administer using the DBMS_JOB package. On the other hand, requiring all jobs to be submitted under a single schema can introduce significant administrative overhead and complexity.

Fortunately, there is a way out of this dilemma. While it is not widely documented (until now), there is a hidden package interface into the job queue, which allows administrators to manipulate jobs that are not their own. This package is called DBMS_IJOB and it is created entirely in the prvtjob.plb script. DBMS_IJOB allows properly authorized users to manipulate any job in the job queue.

The following procedure uses DBMS_IJOB.BROKEN to set or unset the broken flag for all jobs in the job queue:

/* Filename on companion disk: 

job3.sql */*
PROCEDURE 

break_all_jobs (set_broken_IN IN BOOLEAN)
IS
   /*
   || Sets the broken flag to TRUE or FALSE for all 
   || jobs in the job queue
   ||
   || Requirements:
   ||
   || SELECT on DBA_JOBS
   || EXECUTE on DBMS_IJOB
   */
BEGIN
   FOR job_rec IN 
            (SELECT job 
               FROM dba_jobs)
   LOOP
      SYS.DBMS_IJOB.BROKEN(job_rec.job, set_broken_IN);
   END LOOP;
END break_all_jobs;

Another useful administrative feature of DBMS_IJOB is the ability to remove other users' jobs from the queue. Again, such activities should typically be done by DBAs and only when necessary. Here is a handy procedure similar to the previous one that will remove all jobs by user, or all jobs if NULL is explicitly passed in for the owner_IN parameter. If no job owner is specified, the procedure removes all jobs owned by the caller.

/* Filename on companion disk: job3.sql */*
PROCEDURE 

remove_all_jobs
   (owner_IN IN VARCHAR2 := USER )
IS
   /*
   || Removes all jobs from the job queue owned by
   || a specific user, defaults to current user
   ||
   || Requirements:
   ||
   || SELECT on DBA_JOBS
   || EXECUTE on DBMS_IJOB
   */
BEGIN
   FOR job_rec IN 
            (SELECT job 
               FROM dba_jobs 
              WHERE priv_user = NVL(owner_IN,priv_user) )
   LOOP
      SYS.DBMS_IJOB.REMOVE(job_rec.job);
   END LOOP;
END remove_all_jobs;

These two procedures may come in handy when manipulation of large numbers of jobs is necessary (e.g., when trying to quiesce an environment with many busy job queue processes).