AutoSys Load Balancing Recommendations for Containerized Agents in OCP and Kubernetes

Products

• Workload Automation Agent

Issue/Introduction

This article explains about machine_method for OCP/Kubernetes Autosys Agents.

• What machine_method is recommended for OCP/Kubernetes AutoSys agents?

• Is cpu_mon fully supported and validated for containerized agents running in OpenShift?

• If OCP agents report 0 available CPU cycles, is that expected behavior due to container resource limits or a known product limitation?

Environment

 

• Container Platform: OpenShift Container Platform (OCP) / Kubernetes

• Agent Release: 24.1

 

Resolution

1. Recommended machine_method for OCP/Kubernetes: It is recommended to use roundrobin (or roundrobin_jobload if max_load capacity configured). Avoid using cpu_mon.

2. Is cpu_mon fully supported and validated for container agents? No, it was designed for physical/VM hosts. The OMCPU job that underpins this method produces unreliable results when running inside containers.

3. Are 0 available CPU cycles expected? Yes — it is the predictable consequence of running the OMCPU job inside a OCP container. It is not a bug but rather a product limitation in containerized deployments.

We recommend avoiding CPU-based load balancing for containerized agent pools. Instead, explicitly configure roundrobin or roundrobin_jobload

Additional Information

The event_demon log shows following message though machine_method is set to roundrobin:

"CAUAJM_I_10531 Machine <xxxxxxx>:Method <roundrobin>. Usage ignored for this method. "

This is an informational message, not an error. It means AutoSys is disregarding machine usage statistics (CPU/load metrics) for that machine, since the ROUNDROBIN load-balancing method — the recommended setting for OCP/Kubernetes-based agents — selects machines by simple rotation rather than resource usage. The machine is not excluded from job distribution; it remains fully active in the round-robin rotation. No action is required.