Device command service
A device management platform for a global PC maker. Schools and districts use it to manage their ChromeOS device fleets, including actions such as restarting or resetting many devices at once.
Client and product names are left out.
What I worked on
- Moved bulk device actions, such as restart and reset, to a Kafka-based queue so that large batches do not overload internal services or the third-party device API. An organisation can have around 250,000 devices.
- Replaced synchronous, multi-threaded command execution with event-driven processing.
- Added retries with exponential backoff and failure handling for the third-party API's rate limits and transient errors.
- Tracked each command through its states in MongoDB, using atomic updates and compound indexes.
- Built paginated APIs that report command status across a fleet.
- Added an MFA check to sensitive device operations.
- Used a distributed lock so scheduled reconciliation and cleanup jobs run on only one Kubernetes pod at a time.
- Wrote unit tests for command state changes and failure cases, and low-level design docs (API contracts, DTOs, schema, sequence flows) before building features.
Technologies
- Java
- Spring Boot
- Kafka
- MongoDB
- Kubernetes