FAQ: Job Resilience on Platform Agent

Introduction

When the Platform Agent is under control of the systemd service, then stopping the Platform Agent service will also kill active OS jobs started by this Agent.

This is caused by the fact that stopping a systemd service -by default- automatically kills all it's CGroup (control group) processes. To verify which processes are currently in the CGroup execute

$ systemctl status scheduler@lnx_agent.service

Solution

To prevent processes from being killed when stopping the systemd service, you need to add option KillMode=process in /etc/systemd/system/scheduler@.service and execute 'sudo systemctl daemon-reload' to effect the change.

...
[Service]
Type=forking
KillMode=process 
...

After this change, when stopping the systemd Platform Agent service, only the main process itself is killed and (child) CGroup processes are left running.

Applies To

RMJ 9 / 2023