We're running a system that is distributed over several servers monitored by Zabbix. Different parts of the service can be dynamically configured to different nodes, and each part has a distinct ID.
Currently a shell script configured as a UserParameter, taking the part ID as an argument, is used to query if a particular part of a service has had any fatal errors. The information is only available on the node responsible for the part. Each part has been configured to the server manually, and monitoring works correctly - until the configuration is changed.
Of course, when different parts are moved to different nodes, the UserParameters start to fail on the original node and are not queried at all on the new node.
It is easy to generate a list of valid arguments for the UserParameter at a given time on the monitored node. If the server could query the valid IDs from an agent, and configure an item for each provided value, monitoring would become a lot easier. Does Zabbix allow an approach like that, or is the only option left changing the monitoring approach?
