Context. Elicitation of requirements in software systems is a complex task. Hence, it requires knowledgeable software architects and other domain experts. Especially in distributed environments such as microservice-based systems, this is the case. Components of the system are developed in isolation and commonly share no prerequisites. Therefore, a precise specification is necessary to ensure the availability and performance of the individual components. Problem. In practice, requirements elicitation of non-functional attributes is often neglected or does not follow a systematic approach. If the elicitation process is performed at all, stakeholders meet in physical group meetings. Here, the presence of domain experts with profound knowledge of the inspected system is required. Such an approach is very time-consuming and can result in high costs. Furthermore, participants are required to cover all essential use cases in a limited number of meetings. Objective. This work proposes Resirio, a requirements chatbot supporting resilience engineers in the elicitation process of resilience-related attributes. Previous works have introduced interactive solutions for requirements elicitation. However, they lack formalization and precise elicitation methodologies. The proposed approach should make the elicitation process more effective for software architects and more accessible to inexperienced requirements engineers. Method. Resirio ensures that requirements are elicited in a systematic approach and stored in template-based ATAM scenarios. We use a CHAZOP-based hazard analysis technique to inspect metrics in traces from Zipkin and Jaeger. Hazards identified by the analysis help to examine stimuli that may lead to deviations from the optimal behavior of the system. With the stimulus as an input, users create ATAM scenarios in a conversation with Resirio. Response and response measure, which are parameters of the scenario, are formalized based on MTL. Result. In a user study with software architects, engineers, and researchers from industry and academia, we evaluated Resirio's usability, effectiveness, and support. We compared the user's interaction with the prototype and examined elicited scenarios during the study. The developed prototype gives novice requirements engineers a foundation for fast requirements elicitation but requires advanced features for expert users to define more precise scenarios. Conclusion. We showcase an interactive solution that enables quick and easy elicitation of resilience-related requirement attributes. Following the systematic CHAZOP-based approach, hazard are identified in traces from Zipkin and Jaeger. In a conversation with Resirio, stakeholders of a microservice-based software system are assisted in refining ATAM scenarios. Results from the user study show that resilience engineers prefer the fast and direct input of Quick Replies to written text.