The agent has to stay in contact with a JBoss ON server. This can either be done by using multicast detection to monitor when its primary JBoss ON server comes online or goes offline or by simply polling the JBoss ON server at intervals to see if the server is online.
These polling methods aren't exclusive; they can both be set, so that the agent can use whatever method is convenient or available to poll the server.
Polling the server allows the agent to stop sending commands and data to the server if the server goes offline and then to resume automatically when the server is back online. If server polling is not enabled on the agent, then the agent always assumes that the server is online and sends its information to the server. If the server goes down, then the agent records repeated connection refused errors, which (if the server is down for a long time) can make the agent log grow very large.
6.15.1. Settings for Polling the JBoss ON Server
The simplest configuration is to set a polling interval for the agent. With this method, the agent simply pings the server at the predefined interval.
Open the agent prompt. For example, if the agent process is already running, the prompt can be opened by re-running the
rhq-agent.sh script with the
setconfig with the
rhq.agent.client.server-polling-interval-msecs setting and a value (in milliseconds). Setting this value to zero (0) or a negative number disables server polling.
> setconfig rhq.agent.client.server-polling-interval-msecs=500
Restart the agent process to load the new configuration.