Програмна Система Моніторингу Розподілених Серверів На Основі Алгоритму Консенсусу Raft

Introduction. The paper focuses on the development of software for a distributed server monitoring system with an integrated Raft consensus mechanism aimed at improving the consistency and reliability of server infrastructure monitoring. Materials and Methods. A system prototype was developed in Go based on a distributed architecture incorporating the Raft consensus algorithm, adaptive metric collection based on the coefficient of variation, and a consensus-based notification delivery mechanism. An experimental study was conducted on clusters comprising 3–7 nodes. System performance was evaluated in terms of throughput, latency, leader re-election time, number of polling operations, and notification delivery completeness. A comparative analysis with Prometheus and Zabbix was also conducted. Results. The system achieved a throughput of 2,800 metrics/s with 3 nodes, a P95 write-path latency of 6.1 ms, and a leader re-election time of 2.4 s. Adaptive metric collection reduced the number of polling operations by 39.8%, while notification delivery completeness reached 100% across the investigated failure scenarios. The comparative analysis based on eight criteria demonstrated differences between the proposed approach and Prometheus and Zabbix, particularly in terms of data consistency and fault tolerance. Conclusions. The proposed approach combines data consistency, adaptive metric collection, and fault-tolerant notification delivery within a distributed server monitoring system, providing a basis for improving the reliability of monitoring processes in distributed server infrastructure.