I have repeatedly run into situations when resources stop getting monitored by HQ when the machine that runs the agent/monitored application gets a new IP address. This is most commonly the case when you have set up the agent and the applications on your laptop and move it between different network locations, e.g. from VPN to corp intranet etc.
The first thing that has to be done which I can somewhat understand is that the agent's data/ directory must be blown away and the agent restarted when a IP address change occurs. This is likely due to the fact the agent always communicates the IP address to the server, as opposed to what you set in the config. Why not just use whatever the user specified and let the server resolve host names if need be?
The other problem which is far more annoying is that once the above step is performed, data from the previously created resources, i.e. server/services still won't be collected. This is reproducible with JVMs monitored via the Spring plugin, using JMX instrumentation as described in the user guide. The only workaround for this I have found is to delete the entire server resource and let it be re-discovered, along with the services attached to it. Of course, this renders any previously collected metric data inaccessible.
The first thing that has to be done which I can somewhat understand is that the agent's data/ directory must be blown away and the agent restarted when a IP address change occurs. This is likely due to the fact the agent always communicates the IP address to the server, as opposed to what you set in the config. Why not just use whatever the user specified and let the server resolve host names if need be?
The other problem which is far more annoying is that once the above step is performed, data from the previously created resources, i.e. server/services still won't be collected. This is reproducible with JVMs monitored via the Spring plugin, using JMX instrumentation as described in the user guide. The only workaround for this I have found is to delete the entire server resource and let it be re-discovered, along with the services attached to it. Of course, this renders any previously collected metric data inaccessible.