This disclosure discloses a fault analysis method and apparatus, a compute device cluster, and a readable storage medium, and relates to the field of computer technologies. The method is applied to a cloud server. The cloud server is configured to run a plurality of services that have a call relationship. The method includes: obtaining a tag of a first service, where the first service is any one of the plurality of services; obtaining link call information of a related service of the first service based on the tag of the first service; and obtaining, based on the link call information, a topology diagram corresponding to the first service, and performing fault analysis on the first service based on the topology diagram corresponding to the first service.