Confluent Software Engineer Onsite Coding Questions

LeetCode #37: Sudoku Solver. Difficulty: Hard. Topics: Array, Hash Table, Backtracking, Matrix. Asked at Confluent in the last 6 months.

LeetCode #36: Valid Sudoku. Difficulty: Medium. Topics: Array, Hash Table, Matrix. Asked at Confluent in the last 6 months.

LeetCode #1797: Design Authentication Manager. Difficulty: Medium. Topics: Hash Table, Linked List, Design, Doubly-Linked List. Asked at Confluent in the last 6 months.

I have an onsite with confluent and one of the coding rounds is as follows per prep call: Type: Practical file operations. Focus: File reading and manipulation. Expectations: Solve practical problems using file APIs. Discuss...

## Problem You are building a health monitor for a distributed system. Nodes send heartbeat pings to a central registry. Implement a class that tracks heartbeats and answers liveness queries: - `ping(node_id, timestamp)` — record that `node_id` was alive at `timestamp`. - `is_alive(node_id, timestamp, timeout)` — return `True` if the node sent a ping within `[timestamp - timeout, timestamp]`. - `get_dead_nodes(timestamp, timeout)` — return all nodes that have NOT pinged within the timeout window. ```python class LivenessMonitor: def ping(self, node_id: str, timestamp: int) -> None: ... def is_alive(self, node_id: str, timestamp: int, timeout: int) -> bool: ... def get_dead_nodes(self, timestamp: int, timeout: int) -> list[str]: ... ``` ``` ping("A", 100) ping("B", 105) ping("A", 150) is_alive("A", 160, 20) -> True # last ping at 150, within 20s is_alive("B", 160, 20) -> False # last ping at 105, 55s ago get_dead_nodes(160, 20) -> ["B"] ``` ## Follow-ups 1. Nodes can re-register after being declared dead. How does your data structure handle that cleanly? 2. How would you scale this to 1 million nodes pinging every second? 3. `get_dead_nodes` is O(n) in your current implementation. Can you make it faster using a sorted structure? 4. What happens if node clocks are skewed by up to 5 seconds relative to the server?