CeresDB: A high-perf, distributed, schema-less and time-series database (Rust)

If you are looking for a production ready distributed store written in Rust. Check out TiKV(https://github.com/tikv/tikv), which was also mentioned in the acknowledge section of the project's README.

There's also a full-featured distributed RDBMS called TiDB built on top of TiKV.

It probably depends on how you define "schema-less". I can't speak for all brands, but in Dynamic Relational, a draft RDBMS idea, if you issue "SELECT madeUpColumn FROM myTable", you'd get nulls for that column (assuming myTable has rows). Depending on settings, the table could also be made up (unused), in which case you'd get zero rows back.

You'd never get errors like "column doesn't exist" or "table does not exist" (unless you add constraints, which would be common as projects mature to ensure data integrity). You can only see if there is non-null data in a named slot. In such a case, there is no formal centralized schema. Any notion of schema would come from a survey of what data exists. Columns and tables are essentially "create on write" (UPDATE and INSERT).

Why not? See for example N1QL used by Couchbase (a NoSQL database).

Couchbase doesn't seem to get a lot of positive feedback but I kind of like it. Flexibility of a NoSQL database combined with a powerful SQL like query language.