EuroGPT Enterprise, an open-source platform based in Europe, prioritizes data privacy. UBICLOUD software engineers faced reliability issues with Hetzner’s AX162 server line, which crashed frequently post-deployment. Despite hardware checks and replacements, the team observed a 16x higher failure rate compared to the previous AX161 model. Investigations into system load, temperature, faulty components, and power consumption suggested the possibility of power limiting as a factor. After collaborating with Hetzner and replacing defective motherboards, the team found improved stability with the latest motherboard versions. Lessons learned included the importance of thorough vetting of new servers and a gradual adoption process.
https://www.ubicloud.com/blog/debugging-hetzner-uncovering-failures-with-powerstat-sensors-and-dmidecode