SRE / DevOps Engineer Interview Experience at Juspay – 2026

This online coding assessment focuses on practical scripting and infrastructure-adjacent logic rather than pure algorithmic puzzles. One question asks for a script that parses a set of server health-check logs and identifies any host that has failed more than a given number of checks within a rolling time window, best solved using a per-host deque of failure timestamps evicted lazily as the window advances.

This technical round covers Kubernetes and monitoring tooling in the context of a high-throughput payments platform - specifically how to set up alerting thresholds for a service processing a very high volume of transactions per second without generating excessive noise during expected traffic spikes. The expected approach alerts on rate-of-change and rolling baseline deviation rather than a single static threshold, since a fixed threshold either misses genuine spikes at low-traffic hours or fires constantly during expected peak windows.

This is a genuinely multi-day hackathon round, originally scoped for two days but reportedly extended into a third with limited additional guidance from the hiring team on how to use the extra time. The task is to design and partially implement a self-healing deployment pipeline component that can detect a failed canary rollout and automatically roll back without manual intervention, working against a partially-scaffolded starter repository. A sound strategy prioritizes getting a working end-to-end rollback path functional on day one before optimizing detection accuracy, since a slow-but-correct pipeline is more defensible to demo than a fast-but-incomplete one.

This broad final round spans Linux, databases, cloud, operating systems, and networking rather than staying within a single topic. Linux questions cover diagnosing a process consuming unexpectedly high memory using tools like ps and pmap; database questions cover indexing strategy for a table with high write volume and how that tradeoff differs from a read-heavy table; networking questions cover how a load balancer's health check interacts with in-flight requests during a rolling deployment, where the expected answer explains the need for connection draining so requests already in progress aren't dropped when a backend is marked unhealthy mid-request.