Off-box agents are consuming too much CPU and memory on your Juniper Apstra controller. In this scenario, how would you solve this problem?

In a Juniper Apstra environment, off-box device agents can run directly on the controller VM. When the number of managed devices grows, these agents can consume significant CPU and memory, impacting the controller's performance. The designed architectural solution for this is to scale horizontally by deploying Apstra worker VMs. These worker VMs form a cluster with the controller, and resource-intensive tasks, such as running off-box agents, can be offloaded from the controller to the workers, thus distributing the load and preserving controller resources.

A. Add more CPU and memory to the Apstra controller VM.

This is vertical scaling and may provide temporary relief, but the recommended Apstra architecture for scaling agent load is horizontal scaling with worker VMs.

C. Use on-box agents instead of off-box agents.

This is a different architectural choice that depends on device support; it doesn't solve the performance issue for an existing off-box agent deployment.

D. Modify the agent profile to consume less resources.

Agent profiles in Apstra are used to define connection parameters (e.g., protocol, credentials), not to manage or throttle agent resource consumption.

1. Juniper Apstra User Guide, "Apstra Cluster". The documentation explains, "For large-scale deployments, you can deploy Apstra worker nodes to which you can offload resource-intensive tasks from the Apstra controller. Off-box device agents are one of the tasks that can be offloaded to worker nodes."

2. Juniper Apstra Deployment Guide, "System Requirements and Sizing". This guide details the specifications for controller and worker VMs, outlining the cluster architecture for scaling deployments.