Skip to content
forked from tidwall/summitdb

In-memory NoSQL database with ACID transactions, Raft consensus, and Redis API

License

Notifications You must be signed in to change notification settings

masomo/summitdb

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

SummitDB

SummitDB is an in-memory, NoSQL key/value database. It persists to disk, uses the Raft consensus algorithm, is ACID compliant, and built on a transactional and strongly-consistent model. It supports custom indexes, geospatial data, JSON documents, and user-defined JS scripting.

Under the hood it utilizes Finn, Redcon, BuntDB, GJSON, and Otto.

Features

Getting started

Getting SummitDB

The easiest way to get SummitDB is to use one of the pre-built release binaries which are available for OSX, Linux, and Windows. Instructions for using these binaries are on the GitHub releases page.

If you want to try the latest version, you can build SummitDB from the master branch.

Building SummitDB

SummitDB can be compiled and used on Linux, OSX, Windows, FreeBSD, ARM (Raspberry PI) and probably others since the codebase is 100% Go. We support both 32 bit and 64 bit systems. Go must be installed on the build machine.

To build simply:

$ make

It's a good idea to install the redis-cli.

$ make redis-cli

To run tests:

$ make test

Running

First start a single-member cluster:

$ ./summitdb-server

This will start the server listening on port 7481 for client and server-to-server communication.

Next, let's set a single key, and then retrieve it:

$ ./redis-cli -p 7481 SET mykey "my value"
OK
$ ./redis-cli -p 7481 GET mykey
"my value"

Adding members:

$ ./summitdb-server -p 7482 -dir data2 -join localhost:7481
$ ./summitdb-server -p 7483 -dir data3 -join localhost:7481

That's it. Now if node1 goes down, node2 and node3 will continue to operate.

Differences between SummitDB and Redis

It may be worth noting that while SummitDB supports many Redis features, it is not a strict Redis clone. Redis has a lot of commands and data types that are not available in SummitDB such as Sets, Hashes, Sorted Sets, and PubSub. SummitDB also has many features that are not available in Redis such as:

  • Ordered key space - SummitDB provides one key space that is a large B-tree. An ordered key space allows for stable paging through keys using the KEYS command. Redis uses an unordered dictionary structure and provides a specialized SCAN command for iterating through keys.
  • Everything a string - SummitDB stores only strings which are exact binary representations of what the user stores. Redis has many internal data types, such as strings, hashes, floats, sets, etc.
  • Raft clusters - SummitDB uses the Raft consensus algorithm to provide high-availablity. Redis provides Master/Slave replication.
  • Javascript - SummitDB uses Javascript for user-defined scripts. Redis uses Lua.
  • Indexes - SummitDB provides an API for indexing the key space. Indexes allow for quickly querying and iterating on values. Redis has specialized data types like Sorted Sets and Hashes which can provide secondary indexing.
  • Spatial indexes - SummitDB provides the ability to create spatial indexes. A spatial index uses an R-tree under the hood, and each index can be up to 20 dimensions. This is useful for geospatial, statistical, time, and range data. Redis has the GEO API which allows for using storing and querying geospatial data using the Geohashes.
  • JSON documents - SummitDB allows for storing JSON documents and indexing fields directly. Redis has Hashes and a JSON parser via Lua.

In-memory with disk persistence

SummitDB store all data in memory. Yet each writable command is appended to a file that is used to rebuild the database if the database needs to be restarted.

This is similar to Redis AOF persistence.

JSON Documents

SummitDB provides the commands JSET, JGET, JDEL for working with json documents.

JSET and JDEL uses the sjson path syntax and JGET uses the gjson path syntax.

Here are some examples:

> JSET user:101 name Tom
OK
> JSET user:101 age 46
OK
> GET user:101
"{\"age\":46,\"name\":\"Tom\"}"
> JGET user:101 age
"46"
> JSET user:101 name.first Tom
OK
> JSET user:101 name.last Anderson
OK
> GET user:101
"{\"age\":46,\"name\":{\"last\":\"Anderson\",\"first\":\"Tom\"}}"
> JDEL user:101 name.last
(integer) 1
> GET user:101
"{\"age\":46,\"name\":{\"first\":\"Tom\"}}"
> JSET user:101 friends.0 Carol
OK
> JSET user:101 friends.1 Andy
OK
> JSET user:101 friends.3 Frank
OK
> GET user:101
"{\"friends\":[\"Carol\",\"Andy\",null,\"Frank\"],\"age\":46,\"name\":{\"first\":\"Tom\"}}"
> JGET user:101 friends.1
"Andy"

JSON Indexes

Indexes can be created on individual fields inside JSON documents.

For example, let's say you have the following documents:

{"name":{"first":"Tom","last":"Johnson"},"age":38}
{"name":{"first":"Janet","last":"Prichard"},"age":47}
{"name":{"first":"Carol","last":"Anderson"},"age":52}
{"name":{"first":"Alan","last":"Cooper"},"age":28}

Create an index:

> SETINDEX last_name user:* JSON name.last

Then add some JSON:

> SET user:1 '{"name":{"first":"Tom","last":"Johnson"},"age":38}'
> SET user:2 '{"name":{"first":"Janet","last":"Prichard"},"age":47}'
> SET user:3 '{"name":{"first":"Carol","last":"Anderson"},"age":52}'
> SET user:4 '{"name":{"first":"Alan","last":"Cooper"},"age":28}'

Query with the ITER command:

> ITER last_name
1) "user:3"
2) "{\"name\":{\"first\":\"Carol\",\"last\":\"Anderson\"},\"age\":52}"
3) "user:4"
4) "{\"name\":{\"first\":\"Alan\",\"last\":\"Cooper\"},\"age\":28}"
5) "user:1"
6) "{\"name\":{\"first\":\"Tom\",\"last\":\"Johnson\"},\"age\":38}"
7) "user:2"
8) "{\"name\":{\"first\":\"Janet\",\"last\":\"Prichard\"},\"age\":47}"

Or perhaps you want to index on age:

> SETINDEX age user:* JSON age
> ITER age
1) "user:4"
2) "{\"name\":{\"first\":\"Alan\",\"last\":\"Cooper\"},\"age\":28}"
3) "user:1"
4) "{\"name\":{\"first\":\"Tom\",\"last\":\"Johnson\"},\"age\":38}"
5) "user:2"
6) "{\"name\":{\"first\":\"Janet\",\"last\":\"Prichard\"},\"age\":47}"
7) "user:3"
8) "{\"name\":{\"first\":\"Carol\",\"last\":\"Anderson\"},\"age\":52}"

It's also possible to multi-index on two fields:

> SETINDEX last_name_age user:* JSON name.last JSON age

For full JSON indexing syntax check out the SETINDEX and ITER commands.

Fencing Tokens

A fencing token is simply a number that increases. It's guaranteed to be consistent across the cluster and can never be deleted or decreased. The value is a 64-bit unsigned integer. The first FENCE call will return "1". This can be useful in applications that need things like distributed locking and preventing race conditions. FENCEGET will read the token without incrementing it.

> FENCE mytoken
"1"
> FENCE mytoken
"2"
> FENCE mytoken
"3"
> FENCEGET mytoken
"3"
> FENCE mytoken
"4"

Built-in Raft Commands

Here are a few commands for monitoring and managing the cluster:

  • RAFTADDPEER addr
    Adds a new member to the Raft cluster
  • RAFTREMOVEPEER addr
    Removes an existing member
  • RAFTPEERS
    Lists known peers and their status
  • RAFTLEADER
    Returns the Raft leader, if known
  • RAFTSNAPSHOT
    Triggers a snapshot operation
  • RAFTSTATE
    Returns the state of the node
  • RAFTSTATS
    Returns information and statistics for the node and cluster

Consistency and Durability

SummitDB is tuned by design for strong consistency and durability. A server shutdown, power event, or kill -9 will not corrupt the state of the cluster or lose data.

All data persists to disk. SummitDB uses an append-only file format that stores for each command in exact order of execution. Each command consists of a one write and one fsync. This provides excellent durability.

Read Consistency

The --consistency param has the following options:

  • low - all nodes accept reads, small risk of stale data
  • medium - only the leader accepts reads, itty-bitty risk of stale data during a leadership change
  • high - only the leader accepts reads, the raft log index is incremented to guarantee no stale data. this is the default

For example, setting the following options:

$ summitdb --consistency high

Provides the highest level of consistency. The default is high.

Leadership Changes

In a Raft cluster only the leader can apply commands. If a command is attempted on a follower you will be presented with the response:

> SET x y
-TRY 127.0.0.1:7481

This means you should try the same command at the specified address.

Hot Backups

SummitDB supports hot-backing up a node. You can retrieve and restore a snapshot of the database to a file using the BACKUP command.

> BACKUP
BULK REPLY OF DATA

Or using an HTTP connection like such:

curl localhost:7481/backup -o backup.db

The backup file format is a series of commands which are stored as RESP Arrays. The command:

SET mykey 123

Is stored on disk as:

"*3\r\n$3\r\nSET\r\n$5\r\nmykey\r\n$3\r\n123\r\n"

To restore a system from a backup, issue each command to the leader. For example, using the nc command you could execute:

$ cat backup.db | nc localhost 7481

Commands

Below is the complete list of commands.

Keys and values
APPEND, BITCOUNT, BITOP, BITPOS, DBSIZE, DECR, DECRBY, DEL, EXISTS, EXPIRE, EXPIREAT, FENCE, FENCEGET, FLUSHDB, GET, GETBIT, GETRANGE, GETSET, INCR, INCRBY, INCRBYFLOAT, KEYS, MGET, MSET, MSETNX, PDEL, PERSIST, PEXPIRE, PEXPIREAT, PTTL, RENAME, RENAMENX, SET, SETBIT, SETRANGE, STRLEN, TTL

JSON JSET, JGET, JDEL

Indexes and iteration
DELINDEX, INDEXES, ITER, RECT, SETINDEX

Transactions
MULTI, EXEC, DISCARD

Scripts
EVAL, EVALRO, EVALSHA, EVALSHARO, SCRIPT LOAD, SCRIPT FLUSH

Raft management
RAFTADDPEER, RAFTREMOVEPEER, RAFTLEADER, RAFTSNAPSHOT, RAFTSTATE, RAFTSTATS

Server
BACKUP

Contact

Josh Baker @tidwall

License

SummitDB source code is available under the MIT License.

About

In-memory NoSQL database with ACID transactions, Raft consensus, and Redis API

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Go 98.4%
  • Shell 1.3%
  • Makefile 0.3%