v0.1 Release Notes#

v0.1 is the initial documentation release for MuduDB. It covers installation, SQL usage, procedure development, package deployment, and architecture.

Documentation synchronized on 2026-10-12 against MuduDB source commit 614413db, using scuptio/mududb.document commit c9451e7 as a reference. These are implementation notes for that checkout, not a claim that all features have been published in a binary release.

New features#

  • SQL CRUD support: CREATE TABLE, INSERT, UPDATE, DELETE, and SELECT with predicates and expressions.

  • Mudu Procedure model: write stored procedures in Rust, AssemblyScript, C, C#, Go, or Python, deploy them as .mpk packages, and invoke them through mcli.

  • MPK packaging: package WebAssembly components, DDL, descriptors, and initialization data into a ZIP-based .mpk archive.

  • Multi-mode server backends: choose between Legacy, IOUring, and Tokio execution modes in mudud.cfg.

  • Per-core worker model: kernel-managed workers with partition-local ownership for reduced cross-core contention.

  • Range partitioning: define global partition rules, bind tables to them, and place partitions on specific workers.

  • Format contracts: formal specifications for page header, tuple binary format, log frame, protocol frame, MPK manifest, server config, file layout, and syscall payload.

  • HTTP management API: install, list, and uninstall applications and query topology through mcli --http-addr.

  • Replication and Raft fault tolerance: three deployment shapes — single node, primary-replica streaming (stream), and Raft consensus (raft); shared replication channel, snapshots and WAL GC, online membership changes (learner two-step), mcli repl tooled failover with epoch fencing, and the GET /mudu/server/replication observability endpoint. See Replication and Raft Fault Tolerance.

Known limitations#

  • The IOUring backend path is KV-first in design orientation; this is a design orientation, not a capability limitation — both SQL and KV run on that backend today.

  • Distributed two-phase commit across multiple workers is not implemented. Cross-worker writes are forwarded and executed remotely, but full atomic commit across workers is not yet supported.

  • Secondary indexes are not supported; queries rely on primary key and range scans.

  • The mudup release installer is a planned design proposal and is not implemented yet; build from source for development.

Migration and compatibility#

  • v0.1 uses format version 1 for all on-disk and wire contracts. Future releases will bump these versions and provide offline migration tools.

  • Configuration files use the v1 TOML schema described in the Server Configuration Contract.