Running Jobs all get killed when an Agent is recycled

Products

• Workload Automation Agent

Issue/Introduction

After creating a service definition to stop/start the Agent on a RedHat 7 Linux system, all jobs that were running when the Agent is stopped are subsequently killed when the Agent is started again.

Cause

Unless otherwise specified in the service definition, the default KillMode is set to control-group. This default setting causes the reported behavior where child processes of a now gone parent process are killed when the process starts again with a new PID.

Environment

Release :

Component :

Resolution

Specify the following KillMode parameter in your service definition:

KillMode=process

With the above, when the Agent is started again with a new PID, the child processes (i.e jobs) that were running will be associated with the new PID in a new process tree and the Agent will be able to continue tracking the jobs without killing them.