Job Scheduler: Solution Walkthrough
High Priority20 min readUpdated November 11, 2026
The Job Scheduler round gives you a small class that decides which job a worker should run next. Jobs can depend on other jobs, multiple workers share the same scheduler, and each worker reports back when it finishes a job.
The code itself is short, but the challenge comes from understanding and correctly handling the scheduling rules. The round has four phases:
- Phase 0: Understand the code. Fill in three blanks in a test file and answer three open questions.
- Phase 1: Fix a bug. A deploy job starts while its build is still running.
- Phase 2: Add a feature. Implement priorities, retries, skipping jobs after a failure, and rejecting dependency cycles.
- Phase 3: Make it fast. Handle 400,000 requests through a single scheduler, with new jobs arriving while it runs and all scheduling rules still enforced.
Each rule is fairly simple on its own. The difficulty comes from how they interact. A retry changes the tie-break order, a failure must reach even jobs that haven't been added yet, and checking for cycles on every new job becomes a performance bottleneck in Phase 3.
Most wrong solutions handle one rule correctly but break another.