Software Engineer - Go/MongoDB (Remote)

About the role

MongoDB Tools Team

Team: MongoDB Tools (Product and Engineering)

Location: Remote

Projects: percona/percona-clustersync-mongodb https://github.com/percona/percona-clustersync-mongodb and percona/percona-backup-mongodb https://github.com/percona/percona-backup-mongodb

ABOUT THE TEAM

The MongoDB Tools Team builds Percona's open source operational tooling for MongoDB. Two projects sit at the center of what we do. Percona ClusterSync for MongoDB (PCSM) clones and continuously replicates data between clusters. Percona Backup for MongoDB (PBM) is a distributed, low-impact backup and restore solution for replica sets and sharded clusters. Both are written in Go, both are Apache 2.0 licensed, and both are built fully in the open.

This role sits primarily on PCSM, which is younger and moving fast, so you will have real influence over how it takes shape. You will also work across into PBM. The two tools share many hard problems: cluster topology, the oplog and change streams, consistency across shards, and performance in very large production clusters. Backup and restore experience is a real advantage here, not just a box to tick.

THE PROJECTS

- PCSM (primary focus): initial data cloning followed by continuous change replication over MongoDB Change Streams, for both replica sets and sharded clusters. Still pre-1.0 and evolving quickly.

- PBM (secondary): consistent backup and restore with point-in-time recovery, using oplog capture to stay consistent across replica sets and sharded clusters, with S3-compatible and filesystem storage. Driven by pbm-agent processes on each node and a pbm CLI. Mature and widely deployed in production.

WHAT YOU WILL DO

PRIMARY, ON PCSM

- The core replication engine: initial collection cloning followed by continuous change capture over MongoDB Change Streams, with correct handling of resume tokens, ordering, and resumability after failures.

- Correctness and fault tolerance at scale: recovering cleanly from network drops, primary elections, and restarts without losing or duplicating changes, and reasoning carefully about the delivery guarantees we can honestly promise.

- Sharded cluster support: replicating across shards, dealing with the realities of chunk migrations and balancer activity, and keeping the target consistent.

- Namespace filtering and automatic index management, plus the edge cases that show up with DDL, TTL, and index differences between source and target.

- Performance and throughput: parallelizing the clone, applying backpressure, and keeping memory and connection use sane against large clusters with great change volume.

- The CLI and HTTP API that drive and observe a sync, and the metrics and logging that let an operator trust what is happening.

ALSO ACROSS PBM

- Consistent backup, restore, and point-in-time recovery across replica sets and sharded clusters, using physical or logical type of the backup.