NDB_REDIS Module

Daniel-Constantin Mierla

Edited by

Daniel-Constantin Mierla

Vicente Hernando

Morten Isaksen

Carsten Bock


Table of Contents

1. Admin Guide
1. Overview
2. Dependencies
2.1. Kamailio Modules
2.2. External Libraries or Applications
3. Parameters
3.1. server (str)
3.2. init_without_redis (integer)
3.3. connect_timeout (int)
3.4. cmd_timeout (int)
3.5. cluster (integer)
3.6. allowed_timeouts (integer)
3.7. disable_time (integer)
3.8. flush_on_reconnect (integer)
3.9. allow_dynamic_nodes (integer)
4. Functions
4.1. redis_cmd(srvname, command, ..., replyid)
4.2. redis_pipe_cmd(srvname, command, ..., replyid)
4.3. redis_execute(srvname)
4.4. redis_free(replyid)

List of Examples

1.1. Set server parameter
1.2. Set init_without_redis parameter
1.3. Set connect_timeout parameter
1.4. Set cmd_timeout parameter
1.5. Set cluster parameter
1.6. Set allowed_timeouts parameter
1.7. Set disable_time parameter
1.8. Set flush_on_reconnect parameter
1.9. Set allow_dynamic_nodes parameter
1.10. redis_cmd usage
1.11. redis_execute usage
1.12. redis_free usage

Chapter 1. Admin Guide

1. Overview

This module provides a connector to interact with REDIS NoSQL Database from configuration file. You can read more about REDIS at http://redis.io.

It can connect to many REDIS servers and store the results in different containers.

2. Dependencies

2.1. Kamailio Modules

The following modules must be loaded before this module:

  • none.

2.2. External Libraries or Applications

The following libraries or applications must be installed before running Kamailio with this module loaded:

  • hiredis - available at https://github.com/antirez/hiredis .

3. Parameters

3.1. server (str)

Specify the details to connect to REDIS server. It takes a list of attribute=value separated by semicolon, the attributes can be name, unix, addr, port, db and pass. Name is a generic identifier to be used with module functions. unix is the path to the unix domain socket provided by redis server. addr and port are the IP address and the port to connect to REDIS server. pass is the server password. unix and (addr, port) are mutually exclusive. If both appear in same server settings unix domain socket is configured. db is the DB number to use (defaults to 0 if not specified).

You can set this parameter many times, in case you want to connect to many REDIS servers, just give different attributes and use the specific server name when querying the REDIS instance.

Default value is NULL.

Example 1.1. Set server parameter

...
modparam("ndb_redis", "server", "name=srvN;addr=127.0.0.1;port=6379;db=1")
modparam("ndb_redis", "server", "name=srvX;addr=127.0.0.2;port=6379;db=4;pass=mypassword")

# Unix domain socket
modparam("ndb_redis", "server", "name=srvY;unix=/tmp/redis.sock;db=3")

# sentinel (for a redis slave)
modparam("ndb_redis", "server", "name=srvZ;sentinel_group=group_name;sentinel_master=0;sentinel=1.2.3.4:26379;sentinel=1.2.3.5:26379")

# sentinel (for a redis master)
modparam("ndb_redis", "server", "name=srvZ;sentinel_group=group_name;sentinel_master=1;sentinel=1.2.3.4:26379;sentinel=1.2.3.5:26379")

...

3.2. init_without_redis (integer)

If set to 1, the module will correctly initialize even if redis is not available at start up.

Default value is 0.

Example 1.2. Set init_without_redis parameter

...
modparam("ndb_redis", "init_without_redis", 1)
...

3.3. connect_timeout (int)

The timeout when connecting to the redis server

Default value is 1000 ms.

Example 1.3. Set connect_timeout parameter

...
modparam("ndb_redis", "connect_timeout", 500)
...

3.4. cmd_timeout (int)

The timeout for each query to the redis server. If the redis server does not reply within the timeout value, the command will fail and kamailio will continue executing the cfg file.

Default value is 1000 ms.

Example 1.4. Set cmd_timeout parameter

...
modparam("ndb_redis", "cmd_timeout", 500)
...

3.5. cluster (integer)

If set to 1, the module will connect to servers indicated in the "MOVED" reply.

The module needs to know all existing REDIS-Nodes at startup. The nodes are searched by the name "ip:port", e.g. if REDIS replies with "MOVED 127.0.0.1:4711", ndb_redis needs to know the databases "127.0.0.1:4711".

Default value is 0 (disabled).

Example 1.5. Set cluster parameter

...
modparam("ndb_redis", "server", "name=127.0.0.1:26001;addr=127.0.0.1;port=26001")
modparam("ndb_redis", "server", "name=127.0.0.1:26004;addr=127.0.0.1;port=26004")
modparam("ndb_redis", "server", "name=127.0.0.1:26008;addr=127.0.0.1;port=26008")
...
modparam("ndb_redis", "cluster", 1)
...

3.6. allowed_timeouts (integer)

If this is set to a non-negative value, it sets the number of consecutive REDIS commands that can fail before temporarily disabling the REDIS server. This is similar to rtpengine_disable_tout parameter from the rtpengine module.

When communicating with a REDIS server, if redis_cmd or redis_execute will fail for more than allowed_timeouts consecutive times, the server will be temporary disabled for a number of seconds configured by the disable_time parameter.

Disabling a server means that further redis_cmd and redis_execute commands will not do anything and return a negative value -2. Messages are also logged when disabling and re-enabling a server.

The number of consecutive fails are counted by each Kamailio process, so when disabling a server this is done just for that process, not globally.

Default value is -1 (disabled).

Example 1.6. Set allowed_timeouts parameter

...
modparam("ndb_redis", "allowed_timeouts", 3)
...

3.7. disable_time (integer)

If allowed_timeouts is set to a non negative value this determines the number of seconds the REDIS server will be disabled

Default value is 0.

Example 1.7. Set disable_time parameter

...
modparam("ndb_redis", "allowed_timeouts", 0)
modparam("ndb_redis", "disable_time", 30)
...

3.8. flush_on_reconnect (integer)

If this is set to a non zero value, a "FLUSHALL" command is issued after reconnecting to a REDIS server, to clear the entire database.

When a command to a REDIS server fails, a reconnection to that server is made, so with this parameter each failed command will result in a flush of the database.

This is useful in scenarios when a REDIS server does not respond to commands, but the commands might have been issued, and the responses lost. If this leaves the data in the db in an uncertain state, a flush might fix any issues that may occur.

Default value is 0 (disabled).

Example 1.8. Set flush_on_reconnect parameter

...
modparam("ndb_redis", "flush_on_reconnect", 1)
...

3.9. allow_dynamic_nodes (integer)

If set to 1, the module will connect to servers indicated in the "MOVED" reply, even if they are not specified at startup.

When cluster is enabled the module supports redirections ("MOVED") replies. Set allow_dynamic_nodes to 1 to avoid listing all the nodes at startup. You can only list one node, e.g. by using a DNS name for the cluster instead of an IP address. The module will add dynamically the new nodes as the MOVED replies are received. Only works if cluster is set to 1.

Default value is 0 (disabled).

Example 1.9. Set allow_dynamic_nodes parameter

...
modparam("ndb_redis", "server", "name=redis_cluster;addr=127.0.0.1;port=26008")
...
modparam("ndb_redis", "cluster", 1)
modparam("ndb_redis", "allow_dynamic_nodes", 1)
...

4. Functions

4.1.  redis_cmd(srvname, command, ..., replyid)

Send a command to REDIS server identified by srvname. The reply will be stored in a local container identified by replyid. All the parameters can be strings with pseudo-variables that are evaluated at runtime.

Minimum required arguments are srvname, command and replyid. Command argument can be separated into several ones using %s token. (See examples) Total number of arguments cannot exceed six.

The reply can be accessed via pseudo-variable $redis(key). The key can be: type - type of the reply (as in hiredis.h); value - the value returned by REDIS server; info - in case of error from REDIS, it will contain an info message.

If reply type is an array (as in hiredis.h), there are other keys available:

  • size - returns number of elements in the array.

  • type[n] - returns the type of the nth element in the array. type - returns array type.

  • value[n] - returns value of the nth element. value - returns null for an array. You need to get each element by index.

In case one of the members of the array is also an array (for example calling SMEMBERS in a MULTI/EXEC transaction), the members of the array can be accessed by adding any of the above keys, after a value[n] key. The first value[n] references the element in the first array, while the next key references an element of the referenced array.

The result type can be compared with $redisd(key) variable to test its value. The key can be: rpl_str, rpl_arr, rpl_int, rpl_err, rpl_sts, rpl_nil.

Example 1.10. redis_cmd usage

...
if(redis_cmd("srvN", "INCR cnt", "r")) {
    # success - the incremented value is in $redis(r=>value)
    xlog("===== $redis(r=>type) * $redis(r=>value)\n");
}

# set a value
redis_cmd("srvN", "SET foo bar", "r");

redis_cmd("srvN", "SET ruri $ru", "r");

# get a value
redis_cmd("srvN", "GET foo", "r");

# same command separated into two arguments:
redis_cmd("srvN", "GET %s", "foo", "r");

# if we have a key with spaces within it:
redis_cmd("srvN", "GET %s", "foo bar", "r");

# several values substitution:
redis_cmd("srvN", "HMGET %s %s %s", "key1", "field1", "field2", "r");


# array example
if(redis_cmd("srvN", "HMGET foo_key field1 field3", "r")) {
    xlog("array size: $redis(r=>size)\n");
    xlog("first values: $redis(r=>value[0]) , $redis(r=>value[1])\n");
}


# array as element of an array example
redis_cmd("srvN", "MULTI", "r1");
redis_cmd("srvN", "SMEMBERS foo", "r2");
if (redis_cmd("srvN", "EXEC", "r")) {
    xlog("array size: $redis(r=>value[0]=>size)\n");
    xlog("first member of the set:$redis(r=>value[0]=>value[0])\n");
    xlog("type of the second member of the set:$redis(r=>value[0]=>type[1])\n");
}
...

4.2.  redis_pipe_cmd(srvname, command, ..., replyid)

Add a command to be sent to REDIS server identified by srvname. All the commands will be stored in a buffer until a call to redis_execute is made. When calling redis_execute the stored commands are sent using the pipelining functionality of redis. The replies will be stored in local containers identified by the replyid of each added command. All the parameters can be strings with pseudo-variables that are evaluated at runtime.

This command is similar in syntax with redis_cmd, the only difference is that it does not send the command but instead appends it to a buffer.

See examples from redis_execute.

Note: Pipelining feature is incompatible with the clustering feature. If cluster parameter is set to 1, this function will log an error and do nothing.

4.3.  redis_execute(srvname)

Sends commands to REDIS server identified by srvname. Commands are added with redis_pipe_cmd function, and will be stored for an existing REDIS server. When this function is called all the commands will be sent in a single message to the REDIS server.

When using redis_cmd together with redis_pipe_cmd it is recommended that a call to redis_execute is made before calling redis_cmd in case there are pipelined commands, otherwise when calling redis_cmd, if pipelined messages exist, a call to redis_execute is made automatically and a warning message is logged.

Note: Pipelining feature is incompatible with the clustering feature. If cluster parameter is set to 1, this function will log an error and do nothing.

Example 1.11. redis_execute usage

...
After several redis command calls:
	redis_pipe_cmd("srvA", "SET foo bar", "r1");

	redis_pipe_cmd("srvB", "SET ruri $ru", "r2");

	redis_pipe_cmd("srvB", "GET foo", "r3");

Send the data and retrieve the results:
	redis_execute("srvA"); //send command to srvA and wait for reply. Store the reply in r1

	redis_execute("srvB"); //send command to srvA and wait for reply. Store the replies in r2 (for SET ruri $ru) and r3 (for GET foo)

Using both redis_cmd and redis_pipe_cmd:
	redis_pipe_cmd("srvA", "SET foo bar", "r1");

	redis_pipe_cmd("srvA", "SET ruri $ru", "r2");

	redis_execute("srvA"); //send commands to srvA and wait for reply. Store replies in r1 and r2

	redis_cmd("srvA", "GET foo", "r3"); //send command, wait for reply and store it in r3


	redis_pipe_cmd("srvA", "SET foo bar", "r1");

	redis_pipe_cmd("srvA", "SET ruri $ru", "r2");

	redis_cmd("srvA", "GET foo", "r3"); //first call redis execute (replies are stored in r1 and r2), log warning and execute redis_cmd

	redis_execute("srvA"); //this does nothing as there are no more pipelined commands. The call is not necessary
...

4.4.  redis_free(replyid)

Frees data in a previous reply from memory. After this function call, accessing to a freed replyid returns null value.

It is not necessary to free a reply to use it again in a new redis_cmd function. When ndb_redis module closes, all pending replies are freed automatically.

Example 1.12. redis_free usage

...
After a redis command call:
	redis_cmd("srvN", "INCR cnt", "r");

free reply data:
	redis_free("r");
...