2.9.1 Supplementary notes on job execution in a cloud environment

(a) Manager host

• Destination for placement

  The manager host can be placed in an on-premise environment and a non-auto-scaling environment of the cloud.

• Cluster configuration

  The manager host supports cluster configuration.

(b) Relay agent

• Destination for placement

  The relay agent can be placed in an on-premise environment and a non-auto-scaling environment of the cloud.

• Cluster configuration

  The relay agent supports cluster configuration.

If the manager host is running on Windows or Linux, you can run the ajsatsetup -m command on the manager host to use the manager host as a relay agent.

In situations shown below, a relay agent is required for a host that is neither a manager host nor a destination agent host.

• If the manager host is running on an OS that is neither Windows nor Linux:

  The OS of the manager host or agent host that a flexible job can specify as an execution host is only Windows or Linux.

  Therefore, if the manager host is running on an OS that is neither Windows nor Linux, a relay agent is necessary.

• If the manager host and the destination agent cannot communicate directly with each other:

  As shown in Figure 2-64, a relay agent must be placed between the manager host and the destination agent to mediate communication.

(c) Destination agent under the load balancer

• Destination for placement

  Place the destination agent in an auto-scaling environment under the load balancer.

  Note that Network Load Balancer and Classic Load Balancer are supported as load balancers (Elastic Load Balancing) for Amazon Web Service, but Application Load Balancer is not supported. In addition, Azure Load Balancer and other load balancers for use with services other than Amazon Web Service are not supported.

• Cluster configuration

  The destination agent does not support cluster configuration. For that reason, by deploying a destination agent in an auto-scaling environment in a cloud, availability in the auto-scaling environment is ensured. Similarly, consider preparing an auto-scaling environment at the remote site as needed.

  Cautionary note

  Scale-in and scale-out actions in an auto-scaling environment in a cloud are managed on the side of the cloud. Thus, the time before startup of the destination agent by scaling-out in an auto-scaling environment finishes and the deletion of destination agents by scaling-in cannot be detected by JP1/AJS3. At that timing, if a flexible job is executed, it might fail. For that reason, regarding the execution of flexible jobs, we recommend using jobs that can be re-executed.