In this video, we're going to learn how Figma managed to scale to 10 million users and the type of challenges that they faced along the way. 

Original Articles 
Part 1: https://www.figma.com/blog/how-figma-scaled-to-multiple-databases/
Part 2: https://www.figma.com/blog/how-figmas-databases-team-lived-to-tell-the-scale/

Music: - ♪ [non copyright music] Lofi Type Beat - Saturday Morning | aesthetic lofi music / Lofiru Link :    [![](https://www.gstatic.com/youtube/img/watch/yt_favicon.png) • [non copyright music] Lofi Type Beat ...](https://www.youtube.com/watch?v=LEMZctXCN7Q&t=0s)

Kiki's Bytes 

Figma scaled from a single PostgreSQL database to a horizontally sharded architecture supporting 10 million users. The journey involved three phases: immediate relief through vertical scaling, read replicas, and connection pooling; medium-term stability via logical partitioning of table groups into dedicated databases; and long-term scalability through horizontal sharding with a custom query router. Key innovations included optimized logical replication by removing indexes during migration, two-phase deployment using PG Bouncer for safe routing changes, logical sharding with PostgreSQL views for testing before physical splits, and a custom Golang DB proxy for intelligent query routing across shards.

How Figma Scaled To 10 Million Users

The animations to explain each concept are very nice and easy to follow. Good watch

The scaling journey from single DB to horizontally sharded architecture is familiar to any team that’s grown past 1M users. The silent killer though isn’t the database scaling itself - it’s managing incidents when your system topology changes.
When you shard 50 tables, suddenly you’ve got cascading failures across shards, inconsistent states that are hard to correlate, and monitoring that breaks. The visibility problem is brutal.
What helped us: building real-time incident context aggregation across shards. When an incident fires, you need to know immediately which shards are affected, what changed in that partition, and how it cascades. That’s the gap between knowing you’re scaled and knowing you’re resilient. Tools like HealOps automatically correlate changes across your sharded topology - saves hours debugging the ‘why’ during an incident.

These animations are really nice, and helped me understand the video really well. Nice work!