Why does HVR kill capture & integrate agents when suspending a job?
When a running job with a capture / integrate agent (action agent/agent plugin) is suspended in the HVR scheduler, HVR seems to kill the agent. Is this correct behavior and why is HVR doing this?
The Capture/Integrate agent as defined with the agent/agent-plugin action is running as an integral part of the replication stream. It is part of the replication job (hence the term plug-in). HVR is capturing its output and status, and acting upon it.
When a running replication job is Suspended, HVR will clean up all parts of this job including a running agent.
This clean-up is done gracefully, by sending a termination signal to the agent process. The agent can therefore be programmed to catch the termination signal and perform a proper rollback when needed.