=encoding utf-8


=head1 Name

lua-resty-redis - Lua redis client driver for the ngx_lua based on the cosocket API



This library is considered production ready.


=head1 Description

This Lua library is a Redis client driver for the ngx_lua nginx module:

https://github.com/openresty/lua-nginx-module/#readme

This Lua library takes advantage of ngx_lua's cosocket API, which ensures
100% nonblocking behavior.

Note that at least L<ngx_lua 0.5.14|https://github.com/chaoslawful/lua-nginx-module/tags> or L<OpenResty 1.2.1.14|http://openresty.org/#Download> is required.


=head1 Synopsis


        # you do not need the following line if you are using
        # the OpenResty bundle:
        lua_package_path "/path/to/lua-resty-redis/lib/?.lua;;";
    
        server {
            location /test {
                content_by_lua '
                    local redis = require "resty.redis"
                    local red = redis:new()
    
                    red:set_timeout(1000) -- 1 sec
    
                    -- or connect to a unix domain socket file listened
                    -- by a redis server:
                    --     local ok, err = red:connect("unix:/path/to/redis.sock")
    
                    local ok, err = red:connect("127.0.0.1", 6379)
                    if not ok then
                        ngx.say("failed to connect: ", err)
                        return
                    end
    
                    ok, err = red:set("dog", "an animal")
                    if not ok then
                        ngx.say("failed to set dog: ", err)
                        return
                    end
    
                    ngx.say("set result: ", ok)
    
                    local res, err = red:get("dog")
                    if not res then
                        ngx.say("failed to get dog: ", err)
                        return
                    end
    
                    if res == ngx.null then
                        ngx.say("dog not found.")
                        return
                    end
    
                    ngx.say("dog: ", res)
    
                    red:init_pipeline()
                    red:set("cat", "Marry")
                    red:set("horse", "Bob")
                    red:get("cat")
                    red:get("horse")
                    local results, err = red:commit_pipeline()
                    if not results then
                        ngx.say("failed to commit the pipelined requests: ", err)
                        return
                    end
    
                    for i, res in ipairs(results) do
                        if type(res) == "table" then
                            if res[1] == false then
                                ngx.say("failed to run command ", i, ": ", res[2])
                            else
                                -- process the table value
                            end
                        else
                            -- process the scalar value
                        end
                    end
    
                    -- put it into the connection pool of size 100,
                    -- with 10 seconds max idle time
                    local ok, err = red:set_keepalive(10000, 100)
                    if not ok then
                        ngx.say("failed to set keepalive: ", err)
                        return
                    end
    
                    -- or just close the connection right away:
                    -- local ok, err = red:close()
                    -- if not ok then
                    --     ngx.say("failed to close: ", err)
                    --     return
                    -- end
                ';
            }
        }




=head1 Methods

All of the Redis commands have their own methods with the same name except all in lower case.

You can find the complete list of Redis commands here:

http://redis.io/commands

You need to check out this Redis command reference to see what Redis command accepts what arguments.

The Redis command arguments can be directly fed into the corresponding method call. For example, the "GET" redis command accepts a single key argument, then you can just call the "get" method like this:


        local res, err = red:get("key")

Similarly, the "LRANGE" redis command accepts threee arguments, then you should call the "lrange" method like this:


        local res, err = red:lrange("nokey", 0, 1)

For example, "SET", "GET", "LRANGE", and "BLPOP" commands correspond to the methods "set", "get", "lrange", and "blpop".

Here are some more examples:


        -- HMGET myhash field1 field2 nofield
        local res, err = red:hmget("myhash", "field1", "field2", "nofield")


        -- HMSET myhash field1 "Hello" field2 "World"
        local res, err = red:hmset("myhash", "field1", "Hello", "field2", "World")

All these command methods returns a single result in success and C<nil> otherwise. In case of errors or failures, it will also return a second value which is a string describing the error.

A Redis "status reply" results in a string typed return value with the "+" prefix stripped.

A Redis "integer reply" results in a Lua number typed return value.

A Redis "error reply" results in a C<false> value I<and> a string describing the error.

A non-nil Redis "bulk reply" results in a Lua string as the return value. A nil bulk reply results in a C<ngx.null> return value.

A non-nil Redis "multi-bulk reply" results in a Lua table holding all the composing values (if any). If any of the composing value is a valid redis error value, then it will be a two element table C<{false, err}>.

A nil multi-bulk reply returns in a C<ngx.null> value.

See http://redis.io/topics/protocol for details regarding various Redis reply types.

In addition to all those redis command methods, the following methods are also provided:




=head2 new

C<syntax: red, err = redis:new()>

Creates a redis object. In case of failures, returns C<nil> and a string describing the error.




=head2 connect

C<syntax: ok, err = red:connect(host, port, options_table?)>

C<syntax: ok, err = red:connect("unix:/path/to/unix.sock", options_table?)>

Attempts to connect to the remote host and port that the redis server is listening to or a local unix domain socket file listened by the redis server.

Before actually resolving the host name and connecting to the remote backend, this method will always look up the connection pool for matched idle connections created by previous calls of this method.

An optional Lua table can be specified as the last argument to this method to specify various connect options:


=over


=item *

C<pool>

Specifies a custom name for the connection pool being used. If omitted, then the connection pool name will be generated from the string template C<< <host>:<port> >> or C<< <unix-socket-path> >>.


=back




=head2 set_timeout

C<syntax: red:set_timeout(time)>

Sets the timeout (in ms) protection for subsequent operations, including the C<connect> method.




=head2 set_keepalive

C<syntax: ok, err = red:set_keepalive(max_idle_timeout, pool_size)>

Puts the current Redis connection immediately into the ngx_lua cosocket connection pool.

You can specify the max idle timeout (in ms) when the connection is in the pool and the maximal size of the pool every nginx worker process.

In case of success, returns C<1>. In case of errors, returns C<nil> with a string describing the error.

Only call this method in the place you would have called the C<close> method instead. Calling this method will immediately turn the current redis object into the C<closed> state. Any subsequent operations other than C<connect()> on the current object will return the C<closed> error.




=head2 get_reused_times

C<syntax: times, err = red:get_reused_times()>

This method returns the (successfully) reused times for the current connection. In case of error, it returns C<nil> and a string describing the error.

If the current connection does not come from the built-in connection pool, then this method always returns C<0>, that is, the connection has never been reused (yet). If the connection comes from the connection pool, then the return value is always non-zero. So this method can also be used to determine if the current connection comes from the pool.




=head2 close

C<syntax: ok, err = red:close()>

Closes the current redis connection and returns the status.

In case of success, returns C<1>. In case of errors, returns C<nil> with a string describing the error.




=head2 init_pipeline

C<syntax: red:init_pipeline()>

C<syntax: red:init_pipeline(n)>

Enable the redis pipelining mode. All subsequent calls to Redis command methods will automatically get cached and will send to the server in one run when the C<commit_pipeline> method is called or get cancelled by calling the C<cancel_pipeline> method.

This method always succeeds.

If the redis object is already in the Redis pipelining mode, then calling this method will discard existing cached Redis queries.

The optional C<n> argument specifies the (approximate) number of commands that are going to add to this pipeline, which can make things a little faster.




=head2 commit_pipeline

C<syntax: results, err = red:commit_pipeline()>

Quits the pipelining mode by committing all the cached Redis queries to the remote server in a single run. All the replies for these queries will be collected automatically and are returned as if a big multi-bulk reply at the highest level.

This method returns C<nil> and a Lua string describing the error upon failures.




=head2 cancel_pipeline

C<syntax: red:cancel_pipeline()>

Quits the pipelining mode by discarding all existing cached Redis commands since the last call to the C<init_pipeline> method.

This method always succeeds.

If the redis object is not in the Redis pipelining mode, then this method is a no-op.




=head2 hmset

C<syntax: red:hmset(myhash, field1, value1, field2, value2, ...)>

C<syntax: red:hmset(myhash, { field1 = value1, field2 = value2, ... })>

Special wrapper for the Redis "hmset" command.

When there are only three arguments (including the "red" object
itself), then the last argument must be a Lua table holding all the field/value pairs.




=head2 array_to_hash

C<syntax: hash = red:array_to_hash(array)>

Auxiliary function that converts an array-like Lua table into a hash-like table.

This method was first introduced in the C<v0.11> release.




=head2 read_reply

C<syntax: res, err = red:read_reply()>

Reading a reply from the redis server. This method is mostly useful for the L<Redis PubE<sol>Sub API|http://redis.io/topics/pubsub/>, for example,


        local cjson = require "cjson"
        local redis = require "resty.redis"
    
        local red = redis:new()
        local red2 = redis:new()
    
        red:set_timeout(1000) -- 1 sec
        red2:set_timeout(1000) -- 1 sec
    
        local ok, err = red:connect("127.0.0.1", 6379)
        if not ok then
            ngx.say("1: failed to connect: ", err)
            return
        end
    
        ok, err = red2:connect("127.0.0.1", 6379)
        if not ok then
            ngx.say("2: failed to connect: ", err)
            return
        end
    
        local res, err = red:subscribe("dog")
        if not res then
            ngx.say("1: failed to subscribe: ", err)
            return
        end
    
        ngx.say("1: subscribe: ", cjson.encode(res))
    
        res, err = red2:publish("dog", "Hello")
        if not res then
            ngx.say("2: failed to publish: ", err)
            return
        end
    
        ngx.say("2: publish: ", cjson.encode(res))
    
        res, err = red:read_reply()
        if not res then
            ngx.say("1: failed to read reply: ", err)
            return
        end
    
        ngx.say("1: receive: ", cjson.encode(res))
    
        red:close()
        red2:close()

Running this example gives the output like this:

    1: subscribe: ["subscribe","dog",1]
    2: publish: 1
    1: receive: ["message","dog","Hello"]

The following class methods are provieded:




=head2 add_commands

C<syntax: hash = redis.add_commands(cmd_name1, cmd_name2, ...)>

I<WARNING> this method is now deprecated since we already do automatic Lua method generation
for any redis commands the user attempts to use and thus we no longer need this.

Adds new redis commands to the C<resty.redis> class. Here is an example:


        local redis = require "resty.redis"
    
        redis.add_commands("foo", "bar")
    
        local red = redis:new()
    
        red:set_timeout(1000) -- 1 sec
    
        local ok, err = red:connect("127.0.0.1", 6379)
        if not ok then
            ngx.say("failed to connect: ", err)
            return
        end
    
        local res, err = red:foo("a")
        if not res then
            ngx.say("failed to foo: ", err)
        end
    
        res, err = red:bar()
        if not res then
            ngx.say("failed to bar: ", err)
        end




=head1 Redis Authentication

Redis uses the C<AUTH> command to do authentication: http://redis.io/commands/auth

There is nothing special for this command as compared to other Redis
commands like C<GET> and C<SET>. So one can just invoke the C<auth> method on your C<resty.redis> instance. Here is an example:


        local redis = require "resty.redis"
        local red = redis:new()
    
        red:set_timeout(1000) -- 1 sec
    
        local ok, err = red:connect("127.0.0.1", 6379)
        if not ok then
            ngx.say("failed to connect: ", err)
            return
        end
    
        local res, err = red:auth("foobared")
        if not res then
            ngx.say("failed to authenticate: ", err)
            return
        end

where we assume that the Redis server is configured with the
password C<foobared> in the C<redis.conf> file:

    requirepass foobared

If the password specified is wrong, then the sample above will output the
following to the HTTP client:

    failed to authenticate: ERR invalid password




=head1 Redis Transactions

This library supports the L<Redis transactions|http://redis.io/topics/transactions/>. Here is an example:


        local cjson = require "cjson"
        local redis = require "resty.redis"
        local red = redis:new()
    
        red:set_timeout(1000) -- 1 sec
    
        local ok, err = red:connect("127.0.0.1", 6379)
        if not ok then
            ngx.say("failed to connect: ", err)
            return
        end
    
        local ok, err = red:multi()
        if not ok then
            ngx.say("failed to run multi: ", err)
            return
        end
        ngx.say("multi ans: ", cjson.encode(ok))
    
        local ans, err = red:set("a", "abc")
        if not ans then
            ngx.say("failed to run sort: ", err)
            return
        end
        ngx.say("set ans: ", cjson.encode(ans))
    
        local ans, err = red:lpop("a")
        if not ans then
            ngx.say("failed to run sort: ", err)
            return
        end
        ngx.say("set ans: ", cjson.encode(ans))
    
        ans, err = red:exec()
        ngx.say("exec ans: ", cjson.encode(ans))
    
        red:close()

Then the output will be

    multi ans: "OK"
    set ans: "QUEUED"
    set ans: "QUEUED"
    exec ans: ["OK",[false,"ERR Operation against a key holding the wrong kind of value"]]




=head1 Load Balancing and Failover

You can trivially implement your own Redis load balancing logic yourself in Lua. Just keep a Lua table of all available Redis backend information (like host name and port numbers) and pick one server according to some rule (like round-robin or key-based hashing) from the Lua table at every request. You can keep track of the current rule state in your own Lua module's data, see https://github.com/openresty/lua-nginx-module/#data-sharing-within-an-nginx-worker

Similarly, you can implement automatic failover logic in Lua at great flexibility.




=head1 Debugging

It is usually convenient to use the L<lua-cjson|http://www.kyne.com.au/~mark/software/lua-cjson.php> library to encode the return values of the redis command methods to JSON. For example,


        local cjson = require "cjson"
        ...
        local res, err = red:mget("h1234", "h5678")
        if res then
            print("res: ", cjson.encode(res))
        end




=head1 Automatic Error Logging

By default the underlying L<ngx_lua|https://github.com/openresty/lua-nginx-module/#readme> module
does error logging when socket errors happen. If you are already doing proper error
handling in your own Lua code, then you are recommended to disable this automatic error logging by turning off L<ngx_lua|https://github.com/openresty/lua-nginx-module/#readme>'s L<lua_socket_log_errors|https://github.com/openresty/lua-nginx-module/#lua_socket_log_errors> directive, that is,


        lua_socket_log_errors off;




=head1 Check List for Issues


=over


=item 1.

Ensure you configure the connection pool size properly in the L<set_keepalive> . Basically if your NGINX handle C<n> concurrent requests and your NGINX has C<m> workers, then the connection pool size should be configured as C<n/m>. For example, if your NGINX usually handles 1000 concurrent requests and you have 10 NGINX workers, then the connection pool size should be 100.

=item 2.

Ensure the backlog setting on the Redis side is large enough. For Redis 2.8+, you can directly tune the C<tcp-backlog> parameter in the C<redis.conf> file (and also tune the kernel parameter C<SOMAXCONN> accordingly at least on Linux). You may also want to tune the C<maxclients> parameter in C<redis.conf>.

=item 3.

Ensure you are not using too short timeout setting in the L<set_timeout> method. If you have to, try redoing the operation upon timeout and turning off L<automatic error logging> (because you are already doing proper error handling in your own Lua code).

=item 4.

If your NGINX worker processes' CPU usage is very high under load, then the NGINX event loop might be blocked by the CPU computation too much. Try sampling a L<C-land on-CPU Flame Graph|https://github.com/agentzh/nginx-systemtap-toolkit#sample-bt> and L<Lua-land on-CPU Flame Graph|https://github.com/agentzh/stapxx#ngx-lj-lua-stacks> for a typical NGINX worker process. You can optimize the CPU-bound things according to these Flame Graphs.

=item 5.

If your NGINX worker processes' CPU usage is very low under load, then the NGINX event loop might be blocked by some blocking system calls (like file IO system calls). You can confirm the issue by running the L<epoll-loop-blocking-distr|https://github.com/agentzh/stapxx#epoll-loop-blocking-distr> tool against a typical NGINX worker process. If it is indeed the case, then you can further sample a L<C-land off-CPU Flame Graph|https://github.com/agentzh/nginx-systemtap-toolkit#sample-bt-off-cpu> for a NGINX worker process to analyze the actual blockers.

=item 6.

If your C<redis-server> process is running near 100% CPU usage, then you should consider scale your Redis backend by multiple nodes or use the L<C-land on-CPU Flame Graph tool|https://github.com/agentzh/nginx-systemtap-toolkit#sample-bt> to analyze the internal bottlenecks within the Redis server process.


=back




=head1 Limitations


=over


=item *

This library cannot be used in code contexts like init_by_luaI<, set_by_lua>, log_by_lua*, and
header_filter_by_lua* where the ngx_lua cosocket API is not available.

=item *

The C<resty.redis> object instance cannot be stored in a Lua variable at the Lua module level,
because it will then be shared by all the concurrent requests handled by the same nginx
worker process (see
https://github.com/openresty/lua-nginx-module/#data-sharing-within-an-nginx-worker ) and
result in bad race conditions when concurrent requests are trying to use the same C<resty.redis> instance
(you would see the "bad request" or "socket busy" error to be returned from the method calls).
You should always initiate C<resty.redis> objects in function local
variables or in the C<ngx.ctx> table. These places all have their own data copies for
each request.


=back




=head1 Installation

If you are using the OpenResty bundle (http://openresty.org ), then
you do not need to do anything because it already includes and enables
lua-resty-redis by default. And you can just use it in your Lua code,
as in


        local redis = require "resty.redis"
        ...

If you are using your own nginx + ngx_lua build, then you need to configure
the lua_package_path directive to add the path of your lua-resty-redis source
tree to ngx_lua's LUA_PATH search path, as in


        # nginx.conf
        http {
            lua_package_path "/path/to/lua-resty-redis/lib/?.lua;;";
            ...
        }

Ensure that the system account running your Nginx ''worker'' proceses have
enough permission to read the C<.lua> file.




=head1 TODO




=head1 Community




=head2 English Mailing List

The L<openresty-en|https://groups.google.com/group/openresty-en> mailing list is for English speakers.




=head2 Chinese Mailing List

The L<openresty|https://groups.google.com/group/openresty> mailing list is for Chinese speakers.




=head1 Bugs and Patches

Please report bugs or submit patches by


=over


=item 1.

creating a ticket on the L<GitHub Issue Tracker|http://github.com/agentzh/lua-resty-redis/issues>,

=item 2.

or posting to the L<OpenResty community>.


=back




=head1 Author

Yichun "agentzh" Zhang (章亦春) E<lt>agentzh@gmail.comE<gt>, CloudFlare Inc.




=head1 Copyright and License

This module is licensed under the BSD license.

Copyright (C) 2012-2016, by Yichun Zhang (agentzh) E<lt>agentzh@gmail.comE<gt>, CloudFlare Inc.

All rights reserved.

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:


=over


=item *

Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.


=back


=over


=item *

Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.


=back

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.




=head1 See Also


=over


=item *

the ngx_lua module: https://github.com/openresty/lua-nginx-module/#readme

=item *

the redis wired protocol specification: http://redis.io/topics/protocol

=item *

the L<lua-resty-memcached|https://github.com/agentzh/lua-resty-memcached> library

=item *

the L<lua-resty-mysql|https://github.com/agentzh/lua-resty-mysql> library


=back