5.3.3 Starting Jobs Without the Dispatcher
Clients who use multiple services to run jobs for one OFSBD database must run the jobs without dispatcher processes. If the client does use dispatchers on each machine, each dispatcher may run each job, which causes duplicate detection results.
-nd
to the
command line after the template ID, as
follows:start_mantas.sh <template id> -nd
Doing so causes the start_mantas.sh
script to execute all jobs in the
template, rather than depending on the dispatcher to run them. The jobs in the template
group run in parallel.
The dispatcher can ensure that it is only running a set number of max jobs at any given time (so if the max is set to 10 and a template has 20 jobs associated to it, only 10 run simultaneously). When running without the dispatcher, you must ensure that the number of jobs running do not overload the system. In the event a job run dies unexpectedly (that is, not through a caught exception but rather a fatal signal), you must manually verify whether any jobs are in the RUN state but do not have a MANTAS process still running, which would mean that the job threw a signal. You must update the status code to ERR to restart the job.
dispatcher
, follow these steps:
- Create the new job and job description through an OFSBD Job Editor.
OFSBD automatically assigns a unique ID to the job when it is created.
- Associate the job to a Job Template Group using the
KDD_JOB_TEMPLATE
table. - Execute the
start_mantas.sh
script with the following parameters:start_mantas.sh <template id> [-sd DD-MON-YYYY] [-ed DD-MON-YYYY] [-nd]
where the optional job parameters
-sd
and-ed
(start date and end date, respectively) are used to constrain the data that an algorithm job pulls back.For example, if these parameters are passed into an Alert Creator job, the Alert Creator considers only matches for a grouping that has a creation date within the range that the parameters specify.
After a job runs successfully in OFSBD, you can no longer copy, edit, or delete the job.