A method (300;600) and system for monitoring distributed applications (for example, running on multiple WAS nodes of a cluster) is proposed. The solution of the invention is based on a self-adaptive resource management infrastructure. Particularly, an authority publishes (306-312) a plurality of rules, each one defining a desired target configuration for a category of subjects. A membership controller on each subject is responsible to assign (315-333) the subject to the respective category; a compliance engine then retrieves and applies (336-351) the rules corresponding to the category of the subject. The resource management infrastructure is used to implement a monitoring on-demand of the distributed application. For this purpose, two rules are defined (603-606) for the WAS nodes having the distributed application in a running condition and in a non-running condition, respectively. Each WAS node having the distributed application in the running condition applies (639-645) the first rule, so as to start the monitoring application; as soon as the distributed application switches to the non-running condition, the WAS node applies (654-657) the second rule, so as to stop the monitoring application automatically.