As part of the Glaber project, a fork of the Zabbix monitoring system was created

Project Glaber develops a fork of the Zabbix monitoring system, aimed at improving efficiency, performance and scalability, and also suitable for creating fault-tolerant configurations that are dynamically executed on multiple servers. Initial project developed as a set of patches to improve the performance of Zabbix, but in April work began on creating a separate fork. Project code spreads licensed under GPLv2.

At heavy loads, Zabbix users face the lack of clustering as such in the free version and problems when it is necessary to store very large amounts of data in the DBMS. Relational DBMSs supported in Zabbix, such as PostgreSQL, MySQL, Oracle and SQLite, are poorly adapted for storing trends for history - sampling by a large number of metrics for half a year will already be “heavy” and you need to optimize the DBMS and queries, build clusters of database servers and etc.

As an output, Glaber implemented the idea of ​​using a specialized DBMS clickhouse, which provides good data compression and very high query processing speed (on the same equipment, you can reduce the load on the CPU and disk system by 20-50 times). In addition to ClickHouse support in Glaber, added various optimizations, such as the use of asynchronous snmp requests, bulk (batch) processing of data from monitoring agents, and the use of nmap to parallelize host availability checks, which made it possible to speed up status polling by more than 100 times. Glaber is also working on support clustering, for which it is planned to use in the future etcd.

Source: opennet.ru