Redis::Cluster::Fast - A fast perl binding for Redis Cluster
use Redis::Cluster::Fast;
Redis::Cluster::Fast::srandom(100);
my $redis = Redis::Cluster::Fast->new(
startup_nodes => [
'localhost:9000',
'localhost:9001',
'localhost:9002',
'localhost:9003',
'localhost:9004',
'localhost:9005',
],
connect_timeout => 0.05,
command_timeout => 0.05,
max_retry_count => 10,
);
$redis->set('test', 123);
# '123'
my $str = $redis->get('test');
$redis->mset('{my}foo', 'hoge', '{my}bar', 'fuga');
# get as array-ref
my $array_ref = $redis->mget('{my}foo', '{my}bar');
# get as array
my @array = $redis->mget('{my}foo', '{my}bar');
$redis->hset('mymap', 'field1', 'Hello');
$redis->hset('mymap', 'field2', 'ByeBye');
# get as hash-ref
my $hash_ref = { $redis->hgetall('mymap') };
# get as hash
my %hash = $redis->hgetall('mymap');
Redis::Cluster::Fast is like Redis::Fast but support Redis Cluster by hiredis-cluster.
To build and use this module you need libevent-dev >= 2.x is installed on your system.
Recommend Redis 6 or higher.
Since Redis 6, it supports new version of Redis serialization protocol, RESP3. This client start to connect using RESP2 and currently it has no option to upgrade all connections to RESP3.
Simple microbenchmark comparing PP and XS. The benchmark script used can be found under examples directory.
Redis::Cluster::Fast is 0.084
Redis::ClusterRider is 0.26
### mset ###
Rate Redis::ClusterRider Redis::Cluster::Fast
Redis::ClusterRider 13245/s -- -34%
Redis::Cluster::Fast 20080/s 52% --
### mget ###
Rate Redis::ClusterRider Redis::Cluster::Fast
Redis::ClusterRider 14641/s -- -40%
Redis::Cluster::Fast 24510/s 67% --
### incr ###
Rate Redis::ClusterRider Redis::Cluster::Fast
Redis::ClusterRider 18367/s -- -44%
Redis::Cluster::Fast 32879/s 79% --
### new and ping ###
Rate Redis::ClusterRider Redis::Cluster::Fast
Redis::ClusterRider 146/s -- -96%
Redis::Cluster::Fast 3941/s 2598% --
hiredis-cluster uses random() to select a node used for requesting cluster topology.
$seed is expected to be an unsigned integer value,
and is used as an argument for srandom().
These are different implementations of Perl's rand and srand. In this client, Perl's Drand01 is also used to determine the destination node for executing a command that is not a cluster command.
Following arguments are available.
Specifies the list of Redis Cluster nodes.
A fractional seconds. (default: 1.0)
Connection timeout to connect to a Redis node.
A fractional seconds. (default: 1.0)
Specifies the timeout value for each read/write event to execute a Redis Command.
A integer value. (default: 5)
The client will retry calling the Redis Command only if it successfully get one of the following error responses. MOVED, ASK, TRYAGAIN, CLUSTERDOWN.
max_retry_count is the maximum number of retries and must be 1 or above.
A fractional value. (default: 1.0)
Specify the number of seconds to treat a series of cluster topology requests as timed out without retrying the operation. At least one operation will be attempted, and the time taken for the initial operation will also be measured.
A value used as boolean. (default: undef)
The client will call CLUSTER SLOTS instead of CLUSTER NODES.
To run a Redis command with arguments.
The command can also be expressed by concatenating the subcommands with underscores.
e.g. cluster_info
It does not support (Sharded) Pub/Sub family of commands and should not be run.
It is recommended to issue disconnect in advance just to be safe when executing fork() after issuing the command.
To run a Redis command in pipeline with arguments and a callback.
The command can also be expressed by concatenating the subcommands with underscores.
Commands issued to the same node are sent and received in pipeline mode.
In pipeline mode, commands are not sent to Redis until run_event_loop, wait_one_response or wait_all_responses is issued.
The callback is executed with two arguments.
The first is the result of the command, and the second is the error message.
$result will be a scalar value or an array reference, and $error will be an undefined value if no errors occur.
Also, $error may contain an error returned from Redis or an error that occurred on the client (e.g. Timeout).
You cannot call any client methods or exceptions inside the callback.
After issuing a command in pipeline mode,
do not execute fork() without issuing disconnect if all callbacks are not executed completely.
$redis->get('test', sub {
my ($result, $error) = @_;
# some operations...
});
This method allows you to issue commands without waiting for their responses. You can then perform a blocking wait for those responses later, if needed.
Executes one iteration of the event loop to process any pending commands that have not yet been sent and any incoming responses from Redis.
If there are events that can be triggered immediately, they will all be processed. In other words, if there are unsent commands, they will be pipelined and sent, and if there are already-received responses, their corresponding callbacks will be executed.
If there are no events that can be triggered immediately: there are neither unsent commands nor any Redis responses available to read,
but unprocessed callbacks remain, then this method will block for up to command_timeout while waiting for a response from Redis.
When a timeout occurs, an error will be propagated to the corresponding callback(s).
The return value can be either 1 for success (e.g., commands sent or responses read), 0 for no callbacks remained, or undef for other errors.
- Be aware that the timeout check will only be triggered when there are neither unsent commands nor Redis responses available to read. If a timeout occurs, all remaining commands on that node will time out as well.
- Internally, this method calls
event_base_loop(..., EVLOOP_ONCE), which performs a single iteration of the event loop. A command will not be fully processed in a single call. - If you need to process multiple commands or wait for all responses, call
this method repeatedly or use
wait_all_responses. - For a simpler, synchronous-like usage where you need at least one response,
refer to
wait_one_response. If you only need to block until all pending commands are processed, seewait_all_responses.
# Queue multiple commands in pipeline mode
$redis->set('key1', 'value1', sub {});
$redis->get('key2', sub {});
# Send commands to Redis without waiting for responses
$redis->run_event_loop();
# Possibly wait for responses
$redis->run_event_loop();
If there are any unexcuted callbacks, it will block until at least one is executed. The return value can be either 1 for success, 0 for no callbacks remained, or undef for other errors.
If there are any unexcuted callbacks, it will block until all of them are executed. The return value can be either 1 for success, 0 for no callbacks remained, or undef for other errors.
Normally you should not call disconnect manually.
If you want to call fork(), disconnect should be call before fork().
It will be blocked until all unexecuted commands are executed, and then it will disconnect.
Normally you should not call connect manually.
If you want to call fork(), connect should be call after fork().
Copyright (C) plainbanana.
This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself.
plainbanana [email protected]