Product Knowledge Base

Space shortcuts

Page tree

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 29 Current »

Introduction

Two types of JobScheduler cluster can be configured, depending on the type of behavior required:

  • Passive Clusters provide backup and fail-over functions
  • Active Clusters.use the JobScheduler's Distributed Orders feature to provide load balancing.

The architecture of these clusters is described in the following diagrams:

Note that both types of cluster can be used to provide:

The configuration of both passive and active clusters is covered in this article.

JobScheduler Cluster Options

  • Two or more JobSchedulers can be operated as a cluster.
  • All JobSchedulers in a cluster have to:
    • be identified by the same JobScheduler ID and 
    • use the same database.
  • Furthermore, each JobScheduler in a cluster has to be started with one of the following cluster options:
    • -exclusive
    • -exclusive -backup
    • -distributed-orders
  • See http://www.sos-berlin.com/doc/en/scheduler.doc/command_line.xml for more information about JobScheduler start options.
  • The following types of JobScheduler clusters can be configured, depending on the cluster start options used.

JobScheduler Passive Clusters

  • We assume a Primary JobScheduler with the start option -exclusive.
  • Backup JobSchedulers will use the start options:
    • -exclusive 
    • -backup 
    • -backup-precedence=n
      • where n is a number.
      • The option -backup-precedence is optional and the number n defines the order in which the Backup JobSchedulers become active.
  • Exclusively the Primary JobScheduler will be active once all the JobSchedulers have been started.
  • All JobSchedulers in the cluster use the same configuration (jobs, job chains, orders, etc.). In particular, the configuration files of the orders must have the same modification date.
  • A Backup JobScheduler will not become active if the Primary JobScheduler terminates in a normal way.
  • If the Primary JobScheduler is aborted or its process is killed (e.g. the server crashes) then the (next) Backup JobScheduler will become active and will be aware of the job states and order states. That is, the 'new' JobScheduler will be aware of whether 
    • jobs or job chains are stopped or active, 
    • job chain nodes are stopped, skipped or active,
    • orders are suspended or active and 
    • what their current state is.
  • The Backup JobScheduler waits 1-2 minutes to be sure that the Primary JobScheduler is dead before it goes active.
    • All starts of jobs or orders that would take place within this time will be lost and they runs on the next scheduled time.
  • If a Backup JobScheduler is active and the primary JobScheduler is restarted then the Backup JobScheduler has to be terminated in order to reactivate the Primary JobScheduler.

See also

Change Management References

T Key Linked Issues Fix Version/s Status P Summary Updated
Refresh

 

JobScheduler Active Clusters

  • All JobSchedulers in this type of cluster use the start options -distributed-orders.
  • All the JobSchedulers in this type of cluster will be active once they all have been started.
  • Each JobScheduler of this cluster will handle its own jobs independently, with the exception of orders for job chains that are configured as distributed.
  • Distributed orders are handled in way that each cluster member can process any step of an order within a job chain. Cluster members use some load balancing logic to determine which JobScheduler instance should process the next job for an order in a job chain.

See also

Change Management References

T Key Linked Issues Fix Version/s Status P Summary Updated
Refresh

 

How can I set the cluster start option?

  • You can set cluster options during installation of the JobScheduler.
  • Edit the ./bin/jobscheduler_environment_variables file.(sh|cmd) as described below to change the cluster options after installation has been completed and the cluster been made operational:

Unix

  • stop the JobScheduler

  • edit ./bin/jobscheduler_environment_variables.sh

     SCHEDULER_CLUSTER_OPTIONS=-exclusive
  • start the JobScheduler

Windows

  • stop the JobScheduler service

  • remove the JobScheduler service

     .\bin\jobscheduler.cmd remove

  • modify .\bin\jobscheduler_environment_variables.cmd

     SET SCHEDULER_CLUSTER_OPTIONS=-exclusive

  • install the JobScheduler service

     .\bin\jobscheduler.cmd install

The Cluster tab in JOC

Only applies for the Classic JOC interface

The information provided in this section only applies to the Classic JOC user interface, which is entering retirement:

FEATURE AVAILABILITY ENDING WITH RELEASE 1.11

The JOC Cockpit user interface replaces JOC:

FEATURE AVAILABILITY STARTING FROM RELEASE 1.11

See the JOC Cockpit Article, which provides a starting point for more information about the operation of Clusters with the JOC Cockpit.

If you open the JobScheduler Operating Center (JOC) of a JobScheduler which is member of a cluster then a Cluster tab will be displayed, showing the cluster members.

Example: Backup Cluster:

 

  • No labels