Turning to what @Peter Schuetze said ...
If the nodes are not virtual machines that you want to control Hudson ( see VMware plugin ), start and stop operations go out of Hudson control. Depending on how you set up your subordinates, Hudson may simply automatically connect when he sees that the node is online, or you may need to make sure that the slave starts something at startup.
You can use the Hudson API (typically HTTP POSTs for URLs on the Hudson host) to tell Hudson that the hosts are going offline ahead of time. This will help avoid crashes that will be killed when the node is down. Check the HTML source on the node page ( http: // hudson / computer / node_name ) to find out what the web interface does for "mark offline" and "disable" operations.
source share