Cloud computing is an online technology to provides computing resources (machines) to end-users on demand for running their applications over the internet. Applications hosted in a cloud computing environment may face fluctuating workloads. To deal with such fluctuating workloads cloud resources are allocated automatically to applications. Allocating cloud resources to applications in an automatic manner is known as Elasticity which can be implemented using auto-scaling. Auto-scaling can be implemented as a reactive or proactive approach. Cloud providers use Virtual Machine based or Container-based virtualization to host applications. Some of the factors that affect the availability of the application are computing resources and users accessing those applications. It is required to allocate/deallocate resources at the right moment, else failing to it can lead to SLA Violation which can result in cloud service user dissatisfaction, negative review for the cloud service provider, etc. During the literature study, it is found that reactive auto-scaling decisions are taken based on CPU utilization threshold. In this paper, we have proposed a reactive auto-scaling algorithm that uses application level (response time) and infrastructure level (CPU utilization) metrics together. This work has been evaluated and validated using our custom microservice-based application. The result shows that our approach improves 4% of SLA achievement and 3% in request processing during a simulation duration of 15 minutes.