Every second, Uber ingests millions of logs from thousands of services across regions, which are analyzed by engineers to identify anomalies and trends. After experiencing organic logging traffic growth, Uber’s ELK-based platform could no longer support their requirements, including frequent schema changes, reduced operational overhead, lower hardware cost, and fast aggregations. As such, they decided to build a new log analysis platform using ClickHouse as their storage technology and an abstraction layer designed to support a schema-agnostic data model. In this post, Chao Wang and Xiaobing describe how the implementation of this platform.