A Raft Implementation for Erlang and Elixir

Ra is a Raft implementation by Team RabbitMQ. It is not tied to RabbitMQ and can be used in any Erlang or Elixir project. It is, however, heavily inspired by and geared towards RabbitMQ needs.

Ra (by virtue of being a Raft implementation) is a library that allows users to implement persistent, fault-tolerant and replicated state machines.

Project Maturity

This library has been extensively tested and is suitable for production use. This means the primary APIs (ra, ra_machine modules) and on disk formats will be backwards-compatible going forwards in line with Semantic Versioning. Care has been taken to version all on-disk data formats to enable frictionless future upgrades.

Status

The following Raft features are implemented:

• Leader election
• Log replication
• Cluster membership changes: one server (member) at a time
• Log compaction (with limitations and RabbitMQ-specific extensions)
• Snapshot installation

Build Status

Supported Erlang/OTP Versions

Ra requires Erlang/OTP 21.3+. Erlang 22+ is highly recommended because of distribution traffic fragmentation.

Quick start

%% First we have to start the Ra application
ra:start(),

%% All servers in a Ra cluster are named processes.
%% Create some Server Ids to pass to the configuration
ErlangNodes = [ra@node1, ra@node2, ra@node3],
ServerIds = [{quick_start, N} || N <- ErlangNodes],

%% start a simple distributed addition state machine with an initial state of 0
ClusterName = quick_start,
{ok, ServersStarted, ServersNotStarted} = ra:start_cluster(ClusterName, {simple, fun erlang:&#39;+&#39;/2, 0}, ServerIds),

%% Add a number to the state machine
%% Simple state machines always return the full state after each operation
{ok, StateMachineResult, LeaderId} = ra:process_command(hd(ServersStarted), 5),

%% use the leader id from the last command result for the next
{ok, 12, LeaderId1} = ra:process_command(LeaderId, 7).

“Simple” state machines like the above can only take you so far. See Ra state machine tutorial for how to write a state machine by implementing the ra_machine behaviour.

Design Goals

• Low footprint: use as few resources as possible, avoid process tree explosion
• Able to run thousands of ra clusters within an Erlang node
• Provide adequate performance for use as a basis for a distributed data service

Use Cases

This library is primarily developed as the foundation for replication layer for replicated queues in a future version of RabbitMQ. The design it aims to replace uses a variant of Chain Based Replication which has two major shortcomings:

• Replication algorithm is linear
• Failure recovery procedure requires expensive topology changes

Documentation

• API docs: https://rabbitmq.github.io/ra/
• How to write a Ra state machine: Ra state machine tutorial
• Design and implementation details: Ra internals guide

Examples

A number of examples can be found in a separate repository.

Configuration

• data_dir:

A directory name where ra will store it’s data.

• wal_data_dir:

A directory name where ra will store it’s WAL (Write Ahead Log) data. If unspecified, data_dir is used.

• wal_max_size_bytes:

The maximum size of the WAL in bytes. Default: 512Mb.

• wal_compute_checksums:

Indicate whether the wal should compute and validate checksums. Default: true

• wal_write_strategy:

  • default:

  The default. Actual write(2) system calls are delayed until a buffer is due to be flushed. Then it writes all the data in a single call then fsyncs. Fastest but incurs some additional memory use.

  • o_sync:

  Like default but will try to open the file with O_SYNC and thus wont need the additional fsync(2) system call. If it fails to open the file with this flag this mode falls back to default

• wal_sync_method:

  • datasync:

  The default. Uses the fdatasync system call after each batch. This avoids flushing file meta-data after each write batch and thus may be slightly faster than sync on some system. When datasync is configured the wal will try to pre allocate the entire WAL file. NB: not all systems support fdatasync. Please consult system documentation and configure it to use sync instead if it is not supported.

  • sync:

  Uses the fsync system call after each batch.

• wal_max_batch_size:

Controls the internal max batch size that the WAL will accept. Higher numbers may result in higher memory use. Default: 32768.

• logger_module:

Allows the configuration of a custom logger module. The default is logger. The module must implement a function of the same signature as logger:log/4 (the variant that takes a format not the variant that takes a fun).

• metrics_key:

Metrics key. The key used to write metrics into the ra_metrics table.

• low_priority_commands_flush_size:

When commands are pipelined using the low priority mode Ra tries to hold them back in favour of normal priority commands. This setting determines the number of low priority commands that are added to the log each flush cycle. Default: 25

[{data_dir, "/tmp/ra-data"},
 {wal_max_size_bytes, 134217728},
 {wal_compute_checksums, true},
 {wal_write_strategy, default},
]

Copyright and License

(c) 2017-2020, VMware Inc or its affiliates.

Double licensed under the ASL2 and MPL1.1. See LICENSE for details.