The central error analyser of the L3 experiment at LEP

We describe the central error analyser (CEA) system developed for the L3 experiment at CERN. The goal was to have a centralized environment to collect error and diagnostic messages coming from the several subsystems of the detector and of the data acquisition system as well as from the slow control. The CEA has been operational during the last three years of the L3 data taking, namely from 1998 till 2000. It reduced significantly the time needed to detect and fix DAQ failures, resulting in an overall efficiency exceeding 90%. We describe how the system handled the error messages and how it managed at a central level the correlation among different error situations. The mechanism to exchange messages among independent processes, running on different platforms like VAX-VMS, HP-Unix and PC-Linux, was based on TCP client-server relationships, while the whole CEA architecture was implemented in C++ on a Linux PC platform and it was based on a relational database (postgresql). (3 refs).