The chronicle paradigm has been used to determine the faults in distributed systems, through the modelled of temporal relationships between observable events, which describe the patterns of the system behaviour. Normally, the diagnostic mechanisms based on chronicles are semi-centralised methods, based on the information from local diagnoses. These models have scalability problems when they are implemented in very large systems. This paper represents the system to be diagnosed as a distributed system composed of several components, and the behaviour of each component is described by its own subchronicles. A chronicle recognition module is assigned to each component, and each local diagnosis result (the recognition of a sub-chronicle) is sent (as an event) to the neighbouring components, so as it can be exploited during the recognition of other sub-chronicles. Additionally, this paper uses continuous query language (CQL), instead of the classic language to define chronicles, to give them more expressiveness.