(Click on images for higher definition view)
Job scheduling can be a critical component of any data processing regime. If payroll is not run or invoices are not generated because the background processes responsible are not run on time, then the fall-out can be significant. It is therefore important to consider whether your chosen job scheduler is intrinsically robust and also whether you can configure it so that it will run even when external forces act against it e.g. when the host hardware on which the scheduling software resides goes down.
cosbatch domains are configured with a main Scheduler Host which is an instance of a cosbatch Server that manages all the object data (data about queues, processors, calendars, for example) and performs all the scheduling of jobs and suites across the cosbatch domain in which it resides. For the purposes of throughput and resilience, more than one cosbatch Server may be installed as part of the domain. Each cosbatch Server is a full implementation of the cosbatch software. The benefit of this is that cosbatch provides a global data model which is mirrored to all peer core hosts, and one or more of these cosbatch Servers may be designated as Control (FailOver) Hosts.
A Control Host will maintain a synchronized copy of the Scheduler Host's database, and in the event of the Scheduler Host going down, will take over the Scheduling responsibility. More than one Control Host may be defined in a cosbatch domain but in the event that more than one such Control Host is defined, then they must be prioritized. The Primary Control Host will seek to take over the scheduling role before the Secondary Control Host and so on. As synchronization of the database can take up significant network bandwidth, OSM normally recommends that no more than two Control Hosts are defined at any one time.
However, you may have other full installations of the cosbatch Server software in the domain which do not act as either Scheduler or Control Hosts. These are called Standard Hosts and are the same in every way other than that they do not maintain a synchronized copy of the database. At any time a Standard Host can be promoted to being a Control Host.