Walk me through how you would debug a production issue where an API endpoint is returning 500 errors for roughly 10% of requests.

role-specific · Mid level · software-engineering

What the interviewer is really asking

Assesses systematic debugging methodology, observability tool familiarity, and calm under production pressure.

What to say

• Start with observability: check error logs and traces for the 10% failing requests — look for patterns (specific user IDs, payloads, regions, time of day).
• Correlate with recent deployments or config changes. If something deployed in the last 24 hours, that's the first suspect.
• Narrow down: reproduce locally with a sample failing payload, add targeted logging, then fix forward with a feature flag or rollback if needed.

What to avoid

• Say 'I'd just restart the servers and see if it goes away.'
• Skip structured investigation and jump straight to code changes without evidence.
• Ignore the 90% success rate as signal — the pattern is valuable.

Example answers

Strong: I filter logs by 500 status, compare the failing requests against succeeding ones, and find the failing ones all share a null user_preferences field. That tells me the bug is in how we handle users who never set preferences — a missing null check after a schema change.

Weak: I'd scale up the servers — probably a capacity issue.