TL;DR
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
rqlite creator Philip O’Toole marks twelve years since he first released the open-source project, which paired SQLite with Raft replication. His retrospective describes how distributed SQLite grew into a varied field spanning consensus-based clusters, backup streaming, mergeable databases and managed cloud products; his account is a first-person history, not an independent survey.
Philip O’Toole, creator of the open-source project rqlite, has published a retrospective marking twelve years since its September 2014 announcement and tracing the emergence of a wider field of distributed SQLite systems. The account describes how projects have taken different approaches to replication, from Raft-based clusters to backup streaming and multi-writer merging, as SQLite moved from a database developers replicated themselves to a feature offered by hosted platforms.
O’Toole says he built rqlite to solve a practical problem: a system he worked on needed a small amount of relational storage, but there was no clear answer to which node should hold it. The project added a Raft consensus log in front of SQLite so writes could be replicated across nodes. He says he published the code on GitHub the weekend he announced it. The first version used the goraft library; in early 2016, rqlite moved to HashiCorp’s Raft implementation, which O’Toole says it still uses.
The central design choice, in O’Toole’s account, was making rqlite a CP system: writes go through a leader, a majority of nodes must acknowledge them, and a minority partition refuses writes. That approach favors consistency and continued agreement over allowing every isolated node to accept writes. Other projects made different trade-offs. Canonical developed dqlite as an embedded database library, while rqlite is a server. Litestream, released by Ben Johnson in 2021, instead streams SQLite’s write-ahead log to cloud storage for recovery, without consensus or a database cluster.
Later projects explored still other models. O’Toole points to cr-sqlite, which uses conflict-free replicated data types to let writers diverge and merge, and libSQL, a SQLite fork used by Turso for a hosted service with edge replicas. His post says Supabase acquired Turso in 2026. It also describes SQLite Cloud as a hosted, Raft-based offering, and Cloudflare’s D1 as a managed SQLite product announced in 2022. These descriptions are part of O’Toole’s retrospective; the post does not provide an independently audited count of projects or market share.
Different Replication Models, Different Needs
The history matters because “distributed SQLite” is not one architecture. Consensus-based systems such as rqlite aim to keep a cluster in agreement and can reject writes when a majority is unavailable. WAL streaming such as Litestream addresses recovery after a machine failure, without promising an immediately available cluster. Merge-based systems prioritize the ability for separate writers to work independently and reconcile later. The choice affects application behavior during outages and network partitions, not just deployment complexity.
That distinction can help teams avoid treating projects as interchangeable. An application that needs a replicated service to accept writes only when a quorum is available has a different problem from a single-machine application seeking a low-cost recovery copy. Local-first software, in turn, may value independent writes even when those writes cannot be made strongly consistent in real time. The trade-off is about the failure mode an application can tolerate.
O’Toole also describes a change in how users encounter the technology: early projects were tools to run or embed, while later offerings made SQLite replication a managed platform feature. That shift can reduce the work of operating a database, but it also means product-specific details—such as replication behavior, availability guarantees and service limits—matter. The retrospective establishes the variety of approaches; it does not compare their performance or reliability under common tests.
Raspberry Pi 5 8GB for self-hosted database server
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
From Open-Source Tool to Hosted Products
O’Toole dates rqlite’s announcement to September 2014 and says it was, to his knowledge, the first open-source system to combine SQLite and Raft. He also gives important qualifications: Expensify had already replicated SQLite internally with a Paxos-style protocol and released Bedrock in 2016, while ActorDB appeared around the same period as rqlite. Those examples mean the underlying idea was not exclusive to rqlite, even if he describes rqlite as an early standalone, general-purpose open-source implementation.
The next developments addressed different needs. Canonical’s dqlite was created for products including LXD and later MicroK8s, as an embedded alternative to running etcd. Litestream’s 2021 release promoted a simpler recovery model for single-machine SQLite deployments. Fly.io later shipped LiteFS, which O’Toole describes as replicating SQLite at the filesystem level using leases from Consul. In 2022, cr-sqlite and libSQL represented divergent technical choices: mergeable replicas in one case and a forked SQLite engine with hosted edge replicas in the other.
O’Toole says coverage of the category often begins around 2021, and presents his post as a correction to that shortened timeline. His account is written by a participant who created one of the projects and worked at InfluxDB; it is useful for dates and stated design motivations, but its perspective should be kept in view when treating claims about precedence or influence.
mini PC for home server Linux Docker
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What the Retrospective Cannot Establish
The post is a first-person account, and several claims—including that rqlite was the first open-source Raft-and-SQLite system—are explicitly framed as O’Toole’s understanding rather than independently verified findings. He acknowledges earlier internal work at Expensify and the near-contemporaneous ActorDB project. The retrospective does not set out a complete inventory of current projects, explain how it defines the category, or provide comparative measures of adoption, reliability, cost or performance.
The source also offers limited detail about individual products’ current technical guarantees and commercial status. It says Supabase acquired Turso in 2026, but does not give the transaction date or terms. It does not quantify how many projects are active, how many customers use them, or how the market is divided between open-source software and managed services. Those points should not be inferred from the project examples alone.
external SSD for database backups and WAL streaming
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Next Questions for SQLite Replication
O’Toole’s retrospective does not announce a new rqlite release or set out a specific next milestone. The practical next step for readers evaluating the field is to compare each system’s documented behavior against the application’s needs: whether it requires a quorum for writes, recovery from a backup, independent writers with later reconciliation, or a provider-managed database. Availability and consistency guarantees should be checked in each project’s current documentation rather than assumed from broad labels such as “distributed SQLite.”
Whether the category continues to consolidate around hosted platforms, or remains a mix of libraries, self-managed tools and specialized services, is not established by the post. Further product releases, changes in ownership and updated documentation may clarify how the different approaches develop; no particular outcome is confirmed in the retrospective.
managed SQLite cloud database service
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is rqlite?
rqlite is an open-source database server that uses SQLite with Raft-based replication. O’Toole says its design routes writes through a leader and requires acknowledgement from a majority of the cluster.
How is rqlite different from Litestream?
According to O’Toole, rqlite uses consensus to replicate writes across a cluster, while Litestream streams SQLite’s write-ahead log to cloud storage for recovery. They address different needs: cluster availability and agreement versus restoring a single-node database.
Does distributed SQLite always use Raft?
No. The projects in the retrospective use different methods. Examples include Raft consensus, WAL streaming, filesystem-level replication and CRDT-based merging. The term describes a broad group of approaches, not one protocol.
What does CP mean in rqlite’s design?
O’Toole describes rqlite as a CP system: it prioritizes consistency and agreement across the cluster. If a network partition leaves a minority without a quorum, that minority refuses writes rather than accepting changes it cannot safely coordinate.
Does the article show which distributed SQLite project is best?
No. The retrospective explains different design goals but does not provide a benchmark or independent comparison of performance, reliability, cost or adoption. The suitable option depends on an application’s consistency, recovery and availability requirements.
Source: hn
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
