Search for: All records

Generative artificial intelligence (AI) technologies, such as ChatGPT have shown promise in solving software engineering problems. However, these technologies have also shown to be susceptible to generating software artifacts that contain quality issues. A systematic characterization of quality issues, such as smells in ChatGPT-generated artifacts can help in providing recommendations for practitioners who use generative AI for container orchestration.We conduct an empirical study with 98 Kubernetes manifests to quantify smells in manifests generated by ChatGPT. Our empirical study shows: (i) 35.8% of the 98 Kubernetes manifests generated include at least one instance of smell; (ii) two types of objects Kubernetes namely, Deployment and Service are impacted by identified smells; and (iii) the most frequently occurring smell is unset CPU and memory requirements. Based on our findings, we recommend practitioners to apply quality assurance activities for ChatGPT-generated Kubernetes manifests prior to using these manifests for container orchestration.