eredis

Non-blocking Redis client with focus on performance and robustness.

Supported Redis features:

• Any command, through eredis:q/2
• Transactions
• Pipelining
• Authentication & multiple dbs
• Pubsub

Example

If you have Redis running on localhost, with default settings, you may copy and paste the following into a shell to try out Eredis:

git clone git://github.com/wooga/eredis.git
cd eredis
./rebar compile
erl -pa ebin/
{ok, C} = eredis:start_link().
{ok, <<"OK">>} = eredis:q(C, ["SET", "foo", "bar"]).
{ok, <<"bar">>} = eredis:q(C, ["GET", "foo"]).

MSET and MGET:

KeyValuePairs = ["key1", "value1", "key2", "value2", "key3", "value3"].
{ok, <<"OK">>} = eredis:q(C, ["MSET" | KeyValuePairs]).
{ok, Values} = eredis:q(C, ["MGET" | ["key1", "key2", "key3"]]).

HASH

HashObj = ["id", "objectId", "message", "message", "receiver", "receiver", "status", "read"].
eredis:q(C, ["HMSET", "key" | HashObj]).
{ok, Values} = eredis:q(C, ["HGETALL", "key"]).

LIST

eredis:q(C, ["LPUSH", "keylist", "value"]).
eredis:q(C, ["RPUSH", "keylist", "value"]).
eredis:q(C, ["LRANGE", "keylist",0,-1]).

Transactions:

{ok, <<"OK">>} = eredis:q(C, ["MULTI"]).
{ok, <<"QUEUED">>} = eredis:q(C, ["SET", "foo", "bar"]).
{ok, <<"QUEUED">>} = eredis:q(C, ["SET", "bar", "baz"]).
{ok, [<<"OK">>, <<"OK">>]} = eredis:q(C, ["EXEC"]).

Pipelining:

P1 = [["SET", a, "1"],
      ["LPUSH", b, "3"],
      ["LPUSH", b, "2"]].
[{ok, <<"OK">>}, {ok, <<"1">>}, {ok, <<"2">>}] = eredis:qp(C, P1).

Pubsub:

1> eredis_sub:sub_example().
received {subscribed,<<"foo">>,<0.34.0>}
{<0.34.0>,<0.37.0>}
2> eredis_sub:pub_example().
received {message,<<"foo">>,<<"bar">>,<0.34.0>}

Pattern Subscribe:

1> eredis_sub:psub_example(). 
received {subscribed,<<"foo*">>,<0.33.0>}
{<0.33.0>,<0.36.0>}
2> eredis_sub:ppub_example().
received {pmessage,<<"foo*">>,<<"foo123">>,<<"bar">>,<0.33.0>}
ok
3> 

EUnit tests:

./rebar eunit

Commands

Eredis has one main function to interact with redis, which is eredis:q(Client::pid(), Command::iolist()). The response will either be {ok, Value::binary() | [binary()]} or {error, Message::binary()}. The value is always the exact value returned by Redis, without any type conversion. If Redis returns a list of values, this list is returned in the exact same order without any type conversion.

To send multiple requests to redis in a batch, aka. pipelining requests, you may use eredis:qp(Client::pid(), [Command::iolist()]). This function returns {ok, [Value::binary()]} where the values are the redis responses in the same order as the commands you provided.

To start the client, use any of the eredis:start_link/0,1,2,3,4,5 functions. They all include sensible defaults. start_link/5 takes the following arguments:

• Host, dns name or ip adress as string
• Port, integer, default is 6379
• Database, integer or 0 for default database
• Password, string or empty string([]) for no password
• Reconnect sleep, integer of milliseconds to sleep between reconnect attempts

Reconnecting on Redis down / network failure / timeout / etc

When Eredis for some reason looses the connection to Redis, Eredis will keep trying to reconnect until a connection is successfully established, which includes the AUTH and SELECT calls. The sleep time between attempts to reconnect can be set in the eredis:start_link/5 call.

As long as the connection is down, Eredis will respond to any request immediately with {error, no_connection} without actually trying to connect. This serves as a kind of circuit breaker and prevents a stampede of clients just waiting for a failed connection attempt or gen_server:call timeout.

Note: If Eredis is starting up and cannot connect, it will fail immediately with {connection_error, Reason}.

Redis sentinel support

Overview

Starting from version 2.4.16 and 2.6.0-rc6 redis shipped with standart monitoring and automatic failover tool called Sentinel. It started as separate process that monitors redis instances and automatically switch to new master if the current one fails. After this all slaves are reconfigured to get data from new master automatically by sentinel. More information is here - http://redis.io/topics/sentinel When working with cluster that uses sentinel, clients should ask sentinel processes about current master instance.

Working with sentinels

To enable sentinel support for eredis app:

Start eredis_sentinel main process under supervisor with list of all sentinels as argument:

eredis_sentinel:start_link([{"host1.lan", 20367}, {"host2.lan", 20367}]).

When starting eredis clients use string sentinel:master_name instead host:

eredis:start_link("sentinel:mymaster", 0).

Port is ignored in this case, but needed as eredis:start_link/1 is a special form used in poolboy integration.

eredis_client process will ask eredis_sentinel about current master for mymaster cluster and connect to it. eredis_sentinel also tracks all clients and in case that master changes it will send notifications to all interested clients.

eredis_sentinel implements algorithm described in "Guidelines for Redis clients with support for Redis Sentinel":http://redis.io/topics/sentinel-clients . If it is unable to discover master for some cluster it return error code describing source of problem:

1. sentinel_unreachable - couldn't connect to any of sentinels
2. master_unknown - sentinels do not know about this cluster name
3. master_unreachable - there are no valid master for this cluster now

Testing sentinel support

eredis_sentinel has testing suite wich uses real redis cluster with sentinel monitoring. So for running these tests you should have be allowed to run redis-server and redis-sentinel executables. Test suite is integrated as part of common eredis eunit test suite. Before start it checks that redis-server and redis-sentinel is installed and prints warning if not. Every test case start with fresh cluster with config files from priv/redis_*.conf at the end of case cluster is shutted down.

Pubsub

Thanks to Dave Peticolas (jdavisp3), eredis supports pubsub. eredis_sub offers a separate client that will forward channel messages from Redis to an Erlang process in a "active-once" pattern similar to gen_tcp sockets. After every message sent, the controlling process must acknowledge receipt using eredis_sub:ack_message/1.

If the controlling process does not process messages fast enough, eredis will queue the messages up to a certain queue size controlled by configuration. When the max size is reached, eredis will either drop messages or crash, also based on configuration.

Subscriptions are managed using eredis_sub:subscribe/2 and eredis_sub:unsubscribe/2. When Redis acknowledges the change in subscription, a message is sent to the controlling process for each channel.

eredis also supports Pattern Subscribe using eredis_sub:psubscribe/2 and eredis_sub:unsubscribe/2. As with normal subscriptions, a message is sent to the controlling process for each channel.

As of v1.0.7 the controlling process will be notified in case of reconnection attempts or failures. See test/eredis_sub_tests for details.

AUTH and SELECT

Eredis also implements the AUTH and SELECT calls for you. When the client is started with something else than default values for password and database, it will issue the AUTH and SELECT commands appropriately, even when reconnecting after a timeout.

Benchmarking

Using basho_bench(https://github.com/basho/basho_bench/) you may benchmark Eredis on your own hardware using the provided config and driver. See priv/basho_bench_driver_eredis.config and src/basho_bench_driver_eredis.erl.

Queueing

Eredis uses the same queueing mechanism as Erldis. eredis:q/2 uses gen_server:call/2 to do a blocking call to the client gen_server. The client will immediately send the request to Redis, add the caller to the queue and reply with noreply. This frees the gen_server up to accept new requests and parse responses as they come on the socket.

When data is received on the socket, we call eredis_parser:parse/2 until it returns a value, we then use gen_server:reply/2 to reply to the first process waiting in the queue.

This queueing mechanism works because Redis guarantees that the response will be in the same order as the requests.

Response parsing

The response parser is the biggest difference between Eredis and other libraries like Erldis, redis-erl and redis_pool. The common approach is to either directly block or use active once to get the first part of the response, then repeatedly use gen_tcp:recv/2 to get more data when needed. Profiling identified this as a bottleneck, in particular for MGET and HMGET.

To be as fast as possible, Eredis takes a different approach. The socket is always set to active once, which will let us receive data fast without blocking the gen_server. The tradeoff is that we must parse partial responses, which makes the parser more complex.

In order to make multibulk responses more efficient, the parser will parse all data available and continue where it left off when more data is available.

Future improvements

When the parser is accumulating data, a new binary is generated for every call to parse/2. This might create binaries that will be reference counted. This could be improved by replacing it with an iolist.

When parsing bulk replies, the parser knows the size of the bulk. If the bulk is big and would come in many chunks, this could improved by having the client explicitly use gen_tcp:recv/2 to fetch the entire bulk at once.

Credits

Although this project is almost a complete rewrite, many patterns are the same as you find in Erldis, most notably the queueing of requests.

create_multibulk/1 and to_binary/1 were taken verbatim from Erldis.