A FastAPI model-serving container keeps getting OOM-killed
Interview insights are lockedSee how often this is asked and which companies asked it with Buncho Pro.Situation. A text-classification API runs on FastAPI behind Gunicorn in a container with a 16 GB memory limit. Since the weekend, Kubernetes restarts the pod several times a day with OOMKilled. Latency also spikes before each restart. Nothing changed in the model, but the team recently raised the number of workers to improve throughput and added a result cache.
Use the evidence to explain the failures, and propose short-term mitigation, a root-cause fix and a long-term serving design.
Scenario data
Logs and configuration from the incident. Use them as evidence in your answer.