High Availability and Load Balancing¶
These recommendations should be taken into consideration prior to installing Private Agent(s) to allow for high availability (active/active) and load balancing. Prior to installing a Private Agent, see the System Requirements for Private Agents.
Private Agents were formerly known as Local Agents.
Installing more than one Agent in an Agent Group automatically allows for high availability (active/active). Installing multiple agents in an Agent Group also automatically allows for load balancing.
All agents installed in a specific Agent Group need to be consistent. Every agent installed within an Agent Group needs to be running on the same operating system. When a new agent is installed in an Agent Group, you must verify the operating system of the existing agents and install the new agent with the same operating system. Every agent installed within an Agent Group needs to be running the same version of Jitterbit. When a new agent is installed in an Agent Group, you must verify the version of Jitterbit running on the existing agents. If the new agent is a newer version of Jitterbit, all existing agents will need to be upgraded to the same newer version.
Agents installed on different networks should be in separate Agent Groups. Care should be taken to make sure they have access to the same resources.
Plugins are assigned via the Management Console to Agent Groups, not to individual agents. There should not be any problems when a different agent picks up an operation that relies on a plugin.
All agents in an Agent Group are synchronized with the Harmony cloud server, providing consistent execution of operations. The Harmony cloud server also sends the run operation messages to the agents, and balances the load between them. For example, if you have an Agent Group with two agents and there are two requests to run an operation that come in at the same time, most likely Agent 1 will pick up the first request and Agent 2 will pick up the second request. All child operations are processed on the same agent unless specifically kicked off from the Jitterbit REST service.
When an agent does not respond to the Harmony cloud platform or notifies of its incapability due to high operation load, all operations will automatically be taken over by other agents in the group. If an operation is running when the agent goes down, that operation will fail, as the operation is not able to communicate its status. All other operations will be kicked off normally.
There is no primary agent, or a way to determine which agent will pick up a certain operation. This makes using local files as sources or targets unreliable, as local files are local to the agent that runs the operation. If an agent group contains multiple agents, a fileshare or FTP is recommended for file storage.
Temporary storage sources and targets point to the temporary storage location on the machine that is hosting the agent running the operation. All operations in an operation chain are run on the same agent, so the temporary storage location will be consistent. If an operation that is not in the same chain looks for a file in temporary storage, and you are running more than one agent in an Agent Group, there is no way to determine whether it will look at the same temporary storage location that was used in another operation.