LiteFS aims to work like running a single-server SQLite application as much as possible. However, since it is a distributed system, some things may work differently than you expect. This page is meant to provide answers for commonly asked questions about LiteFS, how it works, and how it differs from other similar database systems.

What is LiteFS used for?

Typically, LiteFS is used for replicating your application data to servers that are close to users. This has a significant impact on reducing latency as application servers can process read requests from their local, on-disk database rather than communicating with a database server over the network.

It can also be used for replicating auxillery databases to a fleet of servers so that changes on the primary are immediately visible to replicas. These databases can include configuration information or they can include supplemental data like geospatial lookups.

What tool(s) does LiteFS replace?

LiteFS works similarly to Postgres or MySQL streaming replication so it can be used as an alternative to those tools when you are using SQLite as your database. SQLite read queries are extremely fast so some users find that they no longer need tools such as Redis or memcached.

What size databases can LiteFS handle?

LiteFS doesn’t have a hard limit for the size of individual SQLite databases. It is typically bound by network and disk I/O as well as the write throughput of the FUSE file system interface. In our testing we typically target SQLite databases that are 10GB or less as that makes up the majority of SQLite databases in real-world applications.

Disk space is usually the limiting factor and we recommend that volumes have at least 20-50% extra disk space that LiteFS can use for storing temporary transaction files. So for a 10GB database, we recommend using a 12-15GB volume.

How many databases can a LiteFS cluster handle?

Each database requires a small amount of memory to track the current state on the primary and each subscribing replica. You should be able to use tens or hundreds of databases on LiteFS without an issue. We have done limited testing on a high number of databases so thousands or hundreds of thousands databases may pose a problem.

What are the tradeoffs of using LiteFS?

LiteFS’ use of FUSE limits the write throughput to about 100 transactions per second so write-heavy applications may not be a good fit. The operating system automatically caches pages in memory so read queries typically don’t see much of a slowdown compared to other file systems. We will be implementing a SQLite VFS implementation in the future which avoids FUSE and that will improve write speeds significantly.

Compared to other replication tools, LiteFS has similar to tradeoffs to other asynchronous streaming replication methods. Asynchronous replication means that there can be a short window of data loss if the primary dies unexpectedly and a write has not been replicated to a replica yet. We will be supporting synchronous replication in the future.

Are LiteFS deploys zero downtime?

Rolling deploys of a LiteFS cluster will not experience any read availability loss but may have a short window of a few milliseconds of write availability loss. LiteFS can handoff the role of primary to another candidate node automatically but the primary changes can take a few milliseconds to propagate to replica nodes so they may redirect to the old primary during this short window.

In the event of a sudden failure of the primary node, a different candidate node will acquire primary status via a distributed lease from Consul. By default, the lease has a time-to-live (TTL) of 10 seconds so you could lose write availability for up to that amount of time. Your database will still be readable by all LiteFS nodes during this time though.

What’s the difference between LiteFS & Litestream?

Litestream is intended as a single-node, disaster recovery tool. If you are only running a single server then it can be a great option. The main tradeoffs of using Litestream are that it cannot replicate data to other live servers and it does not support automatic failover.

LiteFS was originally split off of Litestream in order to keep Litestream as a simple disaster recovery tool. LiteFS improves upon Litestream by adding live replication to replica servers and it provides failover by using Consul for distributed leases.

When making a decision between the two tools, you’ll typically choose Litestream for single-server deployments and LiteFS for multi-server deployments.