SQLite in production is not the meme it used to be

Three years ago saying 'we run SQLite in production' was a mild confession. Today, given Litestream, LiteFS, Turso, and the read performance on modern NVMe, I think the stigma is mostly outdated.

Our current setup: SQLite + Litestream to S3 for continuous backup. Single-process writes (never been a bottleneck for us), reads are blazing. Operational complexity is near zero — no connection pooling, no separate DB server, zero network round trips for queries.

The cases where it still doesn't work: write-heavy workloads, multi-node write requirements, or very large databases where WAL checkpointing becomes painful. For most startups doing <10k writes/day, you're probably fine.

Links: https://litestream.io | https://turso.tech | https://fly.io/blog/all-in-on-sqlite-litestream/