Suppose your main REST endpoint that aggregates data from multiple services has started timing out under increased load. How would you systematically identify the root cause and stabilize the endpoint in the short term while planning a longer-term fix?

Question

VirtualInterview.AI · Accepted Answer

This actually happened on an API I owned that fanned out to 5 backend services and suddenly started hitting 30–40% timeouts after a traffic spike. My first step was to get visibility: enable detailed tracing (we used OpenTelemetry), add correlation IDs, and break down latency by downstream call. Very quickly we saw one service consistently spiking from ~80ms to 1.5–2s under load.

Short term, we added aggressive timeouts, circuit breakers, and degraded responses: if that service was slow, we returned partial data with a clear flag instead of waiting. That alone cut timeouts from 40% to under 5% the same day.

In parallel, we worked with the owning team to fix their N+1 queries and add an index, which brought their p95 down by ~70%. Longer term, we redesigned the endpoint to precompute some aggregates via a Kafka consumer and cache results in Redis, so at peak we served 90% of traffic from cache with sub-50ms latency.

Sample Answer

Keywords