Introduction

Agents can be clustered for high availability and scalability. The motivation for clustering includes:

• High availability: Should a server node in the scheduling environment fail then jobs are executed with the Agents on remaining server nodes.
• Performance: A single Agent can execute thousands of jobs in parallel tasks, however, for a number of use cases it is preferable to distribute server load. This applies particularly to applications that require horizontal scaling across a number of server nodes instead of vertical scaling on a more powerful single server node.

Agent clustering is subject to the agreements of the JS7 - License and is available to commercial license holders. Cluster Agents receive license information from a Controller and do not require individual license keys, see JS7 - How to apply a JS7 License Key.

Use of Standalone Agents is available to Open Source License holders and to commercial license holders, for details see JS7 - Management of Standalone Agents.

Development Status:

• Subagent Cluster:, see  JS-1954 - Getting issue details... STATUS
  FEATURE AVAILABILITY STARTING FROM RELEASE 2.2.0
• Director Cluster see  JS-1955 - Getting issue details... STATUS
  FEATURE AVAILABILITY STARTING FROM RELEASE 2.3.0

Architecture

The architecture is explained with JS7 - Agent Cluster.

The architecture includes to consider different aspects of clustering:

• The Operational Layer covers installation of Cluster Agents that is no different from installation of Standalone Agents. In addition it covers registration of Agents as members of an Agent Cluster.
• The Functional Layer is situated on top of the Operational Layer and defines a number of logical Subagent Clusters.

Operational Layer

The architecture includes the clustering of Agents in the role of Director Agents for restart capabilities and the clustering of Agents acting as Subagents for high availability and scalability purposes.

• The components involved in an Agent Cluster are represented as one Agent to the Controller.
• A Controller can manage any number of Agents from different Agent Clusters or as a number of Standalone Agents.

Functional Layer

The architecture allows to define any number of Subagent Clusters that include a Selection and Scheduling Mode of Subagents: 

• Any number of Subagents can be selected to be part of a Subagent Cluster.
• The selection includes to specify the Scheduling Mode being one of fixed-priority or round-robin:
  • fixed-priority: execute jobs with the first Subagent and switch to the next Subagent only if the first Subagent becomes unavailable (active-passive cluster).
  • round-robin: execute each next job on the next Subagent (active-active cluster).

Manage Agent Clusters

The JS7 - Agent Installation On Premises and JS7 - Agent Installation for Docker Containers are the same for Cluster Agents and for Standalone Agents.

The  icon in the main menu offers to select the Manage Controllers/Agents view:

This brings forward the following view:

• The view is grouped in Controllers (indicated by testsuite and standalone).
• For each Controller separate lists of Standalone Agents and Cluster Agents are displayed.
  
  

For Cluster Agents the following excerpt from the above screenshot is used:

• In a first position the Primary Director Agent is available. Optionally a Secondary Director Agent can be available.
• Below the Director Agent entry the list of Subagents is displayed.
  
  

Add Agent Cluster

The Agent Cluster is situated in the operational layer and includes specification of Director Agents and Subagents.

To add an Agent Cluster users can start from the Controller's action menu to which the Agent Cluster should be added:

This brings forward the following popup window:

Explanation:

• The first group of input fields specifies the Agent Cluster:
  • Agent ID: A unique identifier of the Agent Cluster. Uniqueness is required for all Standalone Agents and Agent Clusters assigned the same Controller. This identifier cannot be changed later on.
  • Agent Cluster Name: The name of the Agent Cluster as displayed in the JOC Cockpit GUI. The name can be changed at any later point in time.
  • Title: An individual description that can be searched for.
  • Alias Name: The same Agent Cluster can appear with a number of alias names. This allows to map use of different Agent Clusters, for example in a production environment, to use of fewer Agent Cluster using alias names in a test environment.
• The second group of input fields specifies the Primary Director Agent:
  • Subagent ID: A unique identifier of the Subagent. Uniqueness is applied for all Director Agents and Subagents in the same Agent Cluster. This identifier cannot be changed later on.
  • Title: An individual description that can be searched for.
  • URL: The protocol HTTP or HTTPS, host name or IP address and port by which the Director Agent is accessible to the Controller.
• The third group of input fields specifies the Secondary Director Agent should Director Agents be clustered.

Note:

• Each Director Agent includes a Subagent that can be used for execution and that is automatically available for use with any Subagent Cluster.

Add Subagent

With the Agent Cluster being added users can then add further Subagents from the Agent Cluster's action menu:

This brings forward the following popup window:

Explanation:

• Subagent ID: A unique identifier of the Subagent. Uniqueness is applied for all Director Agents and Subagents in the same Agent Cluster. This identifier cannot be changed later on.
• Title: An individual description that can be searched for.
• URL: The protocol HTTP or HTTPS, host name or IP address and port by which the Subagent is accessible to the Director Agent.

Enable/Disable Subagent 

Subagents can be disabled and enabled. They are enabled by default.

• When a Subagent is disabled then it is no longer considered for job execution.
  • At the point in time of disabling the Subagent will continue to complete any tasks running for jobs.
  • No new tasks are added to a disabled Subagent.
  • The remaining Subagent(s) in the cluster will take on the load.
• When a disabled Subagent is enabled then it is automatically considered for next job executions.
• The assignment of Subagent Clusters that include a disabled Subagent to jobs in a workflow remains unchanged.
  • For clusters that include more than one Subagent the remaining Subagents will execute the jobs.
  • For Subagent Clusters that hold a single disabled Subagent any orders will wait for the Subagent to become available and enabled. When the Subagent is up and running, for example after a restart, and is enabled then orders will automatically continue in the workflow.

Disabling of Subagents can be used for example to manage maintenance windows on server nodes.

Delete Subagent

Subagents can be deleted only if no jobs are assigned the Subagent Cluster in question.

• Users can apply the Search & Replace function for Agent assignments in a workflow that is available from the JS7 - Workflow Editor.
• Users can apply the JS7 - Inventory Search function to replace Agent assignments across workflows.
• Modified workflows have to be deployed to a Controller to make such changes effective.

Deploy Agent Cluster

With the configuration of an Agent Cluster being completed the cluster can be deployed to its Controller.

• An Agent Cluster and its Subagent Clusters can be used only after deployment to a Controller.
• Changes to an Agent Cluster, for example if the Director Agent's URL is changed or if a Subagent is added or dropped, require to be deployed.
• Deployment of an Agent Cluster includes to deploy any Subagent Clusters configured for the Agent Cluster. However, Subagent Clusters can be deployed individually, see chapter: Deploy Subagent Cluster.

Manage Subagent Clusters

Subagent Clusters are situated in the Functional Layer and include specification of logical clusters that make use of Subagents.

To add a Subagent Cluster users can start from the Agent Cluster's action menu to which the Subagent Cluster should be added:

This brings forward the following view:

• The view displays the list of available Subagent Clusters.
• The indicated status signals that the Subagent Cluster has been deployed or has not been deployed.
• The indicated synchronization status signals that the definition of the Subagent Cluster in the JS7 inventory and the version deployed to the Controller are in sync.
  
  

The  buttons in the right upper corner allow to toggle between the list view and the card view of Subagent Clusters:

Bulk Operations on Subagent Clusters

For the list view and card view of Subagent Clusters when the checkboxes available with each Subagent Cluster are checked then buttons for the following bulk operations become available:

• Revoke: The Subagent Cluster is revoked from the Controller, i.e. the deployment is undone. The definition of the cluster remains in place and can later on be deployed once again.
• Delete: The Subagent Cluster is deleted from the Controller and from the JS7 inventory. This operation cannot be undone.
• Deploy: The Subagent Cluster is forwarded to the Controller and can be assigned to jobs.
  
  

Add Subagent Cluster

When adding Subagents the JOC Cockpit can be instructed to automatically create a Subagent Cluster for the single Subagent.

To add a Subagent Cluster the Create Subagent Cluster button can be used.

This brings up the following popup window:

• In the left panel the list of Subagents is available and can be dragged & dropped to the middle panel.
• The middle panel holds a drop area to which a Subagent can be dragged & dropped.
  
  

With the first Subagent being added to the cluster the middle panel offers additional drag & drop areas:

• The next Subagent can be dropped to the area above the existing Subagent to specify its role as the first Subagent used with the fixed-priority scheduling mode.
  • In this mode the first Subagent is used and only if it becomes unavailable then the next Subagent will be used.
• The next Subagent can be dropped to the area below the existing Subagent to specify its roles as the next Subagent used with the fixed-priority scheduling mode.
  • In this mode the next Subagent is used only if the first Subagent becomes unavailable.
• The next Subagent can be dropped to the area right to the existing Subagent to specify use of the round-robin scheduling mode.
  • In this mode each next task is executed with the next Subagent.
    
    

If the next Subagent is dropped to the area below the existing Subagent then a cluster with fixed-priority scheduling mode is created:

If the next Subagent is dropped to the area right to the existing Subagent then a cluster with round-robin scheduling mode is created:

Deploy Subagent Cluster

With the configuration of a Subagent Cluster being completed the cluster can be deployed to its Controller.

• A Subagent Cluster can be used only after deployment to a Controller.
• Changes to a Subagent Cluster require to be deployed, for example if a Subagent is added or removed from a cluster.
• Deployment of Subagent Clusters can be performed individually and from bulk operations, see chapter: Bulk Operations on Subagent Clusters.
• When deploying an Agent Cluster then this includes to deploy any available Subagent Clusters, see chapter: Deploy Agent Cluster.

The operation to deploy a Subagent Cluster is available from the Deploy button in the right upper corner:

Delete Subagent Cluster

Subagent Clusters can be deleted only if no jobs are assigned the Subagent Cluster in question.

• Users can apply the Search & Replace function for Agent assignments in a workflow that is available from the JS7 - Workflow Editor.
• Users can apply the JS7 - Inventory Search function to replace Agent assignments across workflows.
• Modified workflows have to be deployed to a Controller to make changes effective.

Assign Subagent Cluster to Jobs

Each Subagent Cluster is individually assigned to a job.

• The property editor offers to select from the list of Agents that can be assigned a job.
  • This includes Standalone Agents.
  • This includes Subagent Clusters that are grouped by Agent Cluster.
• Assignment of Subagent Clusters can look like this:
  
  

Further Resources