Kafka at Linkedin processes over 3 Trillion messages a day with over 2000 kafka brokers. At such a scale, maintaining balanced workload on kafka clusters as they go through irregular traffic patterns and hardware failures is a daunting task. SREs at Linkedin expend significant time and effort in handling these curveballs and making sure the hardware resources are utilized evenly, which made it quite evident that intelligent automation was crucial to scale any further. This talk outlines Linkedin’s approach towards solving this problem with the help of Kafka Cruise Control.