Using Mnesia in an Elixir Application

The post presents Mnesia as an OTP-built DBMS available to Elixir and Erlang without extra dependencies or separate services. It focuses on the design trade-offs that make Mnesia fit soft real-time key/value access, complex queries, and replicated tables with fault tolerance.

Reading notes

#

• Mnesia is built into OTP and can be used directly from Elixir or Erlang without installing extra dependencies or running a separate database system.
• It was designed for telecommunications workloads that need fast key/value lookup, complex queries, and high fault tolerance.
• Because it runs in the same memory space as the application, it avoids network overhead and can deliver soft real-time access.
• Mnesia stores Erlang data types natively, so read and write operations do not require marshalling and unmarshalling.
• For complex queries, the post mentions Query List Comprehensions, Match Specifications, and table indexes.
• Mnesia replicates tables across cluster nodes, and transactions write to all configured replicas.
• If some nodes are unavailable during a write, the available replicas are updated and the missing ones are synchronized when they return.
• In CAP terms, the post describes Mnesia as closer to a CP database because it propagates commits across replicas and does not support eventual consistency.
• The recommendation for using Mnesia depends on the application’s requirements and trade-offs.
• The author says Mnesia worked well in production for soft real-time session data that lived only for the duration of a user session.
• For large datasets or cases that do not need soft real-time latency, the post suggests traditional databases such as MySQL or Postgres.
• For needs similar to Redis or Memcached, the post suggests considering Mnesia because it is built into OTP.
• The hands-on section builds a simple banking app with Amnesia as an Elixir wrapper around the Mnesia API.
• The sample app supports creating accounts, transferring money, fetching account details, depositing funds, withdrawing funds, and searching for accounts with low balances.
• The database schema defines a single Account table with an auto-incrementing id, first name, last name, balance, and an index on balance.
• The tutorial creates the database with mix amnesia.create -d Database --disk, which produces a Mnesia.nonode@nohost directory for persisted data.
• The Accounts module creates accounts inside transactions and reads accounts back by id.
• The transfer, deposit, and withdraw functions update balances inside transactions, and the transfer logic checks for sufficient funds and valid source and destination accounts.
• The low-balance query uses Account.where(balance < min_balance) and Amnesia.Selection.values().
• The sample IEx session shows seeded accounts, a transfer from one account to another, a low-balance query, and persisted data across sessions.