Conference humane::scheduler

1127.0. "Job dependency problem" by HTSC04::WILLIAMCHAN () Wed Jul 03 1996 06:41

Hello,

	Customer is running Scheduler V2.1B-2 and encountered job dependency 
	problem. I've been to customer site and collected some informations
	from it:

	Problem: Customer running more than 300 jobs with dependency related;
	while he found that some of the jobs running repeatly but not to the
	schedule. e.g. A-B-C-D-E-F..... Supposingly, job A run first then
	if A completed successfully, B runs and C runs, so on... But he found
	that the sequence of job submission will out of order. B, C or D will
	run automatically despite of the job A.

	He also told me he had the following foundings:

	There were total three groups of VAXes' running Scheduler:
        TESTBED, DEVELOPMENT, & PRODUCTION; all this three VAXes are
        configured with homogenous dual-host cluster; running same version
        of VMS and Scheduler, moreover the job databases are the same.

	The major difference is only one node is installed and running the
        Scheduler on DEVELOPMENT's & TESTBED's cluster whilst both nodes
        are installed and running Scheduler on PRODUCTION cluster.

        So far the job dispatch problem was found on the PRODUCTION
        cluster; therefore, it is believed that there should be some
        clue related to the clustering.

	There was time difference within the cluster but just 4-5 seconds.

	The job profile for job A is "Next run time" set to "27-Jun-96 19:30",
	"schedule interval" set to "D 19:30"; job B and the rest of jobs' 
	"Next run time" all set to "00:05:00" and "schedule interval" set to
	"D 00:05".
	I can't justify the above setting is right or wrong, therefore I ask
	for any input.
	The version they using is V2.1B-2 , VMS V5.5-2.

	The mentioned symptom started from 03-Jun, 04-Jun, 06-Jun, 10-Jun,
	11-Jun, 12-Jun, 13-Jun, 14-Jun, 15-Jun, 18-21-Jun, after they perform
	the DB_compress and reboot the whole cluster, it was okay for two days,
	but happened again on 24, 25....of Jun.

	Customer just want to know is it due to the cluster synchronizatio 
	problem or their job schedule parameter settings. Simply speaking,
	they want to know how to create the dependency jobs with following
	criteria:

	1. Job A with schedule such as daily running at specific time.
	2. Job B and job Z... run after the previous dependent job completed 
	successfully. 
	3. If he want job B and job Z.... run immediately after the completion
	of previous dependent job, what can he set in the job profile?
	4. If he want job B and job Z.... run with control such as 2 minutes
	later after the completion of previous dependent job, what can he set
	in the job profile?


	Thanks very much for any input.

Best regards,
William-TK