This module provides a DB APIv1 connector for the Redis server (https://www.redis.io).

It can be used as a replacement for other database modules such as db_mysql and db_postgres. Not all the specs of DB APIv1 are implemented, thus the usage of this module might be restricted to specific modules. Also, for proper performance, this module needs particular configuration tailored to the modules that make use of it.

Since Redis does not provide a schema by itself, db_redis ships with schema files. The path to these has to be defined using the module parameter "schema_path". The schema definition is defined in one file per table, such that the file name corresponds to the table name, and each file is composed of a comma-separated list of column definitions in the format <column-name>/<type>[,<column-name>/<type> ...] in one line, followed by a line holding the table version.

Example definition for the "location" table (from the usrloc module):

username/string,domain/string,contact/string,received/string,path/string,expires/timestamp,q/double,callid/string,cseq/int,last_modified/timestamp,flags/int,cflags/int,user_agent/string,socket/string,methods/int,ruid/string,reg_id/int,instance/string,server_id/int,connection_id/int,keepalive/int,partition/int
8

Because Redis is a key-value store, it requires unique keys. This means that tables and rows from a relational SQL database, e.g. from MySQL, can not be ported one a 1:1 basis to Redis.

For instance, usrloc relies on a key of "username@domain", but in order to store multiple contacts per AoR, it cannot be constrained to uniqueness. To work around this, db_redis supports mapping sets in such a way as to, in the case of the usrloc module, have a set with a key of "username@domain" and its entries being unique keys per contact based on the ruid of a contact. Thus, one contact in usrloc consists of a unique key "location:entry::example-ruid-1" being a hash with the columns like username, domain, contact, path etc. In addition, this unique key is stored in a set "location:usrdom::exampleuser:exampledomain.org". When usrloc does a lookup based on "username@domain", db_redis figures out via the keys/values the query constructed by usrloc to look for the final entry key in the mapping set first. It then query the actual entries from there, avoiding full table scans. For usrloc, the same holds true for expired contacts, requiring a different kind of mapping. There is a certain balance of read performance vs. write performance to consider, because inserts and deletes also have to maintain the mappings, though this yields much faster selects. The mappings can be freely defined, so even though other kamailio modules don't require a specific mapping to be in place for proper performance, mappings could be defined for external applications to read faster (for instance letting the acc module also write mappings besides the actual records for billing systems to correlate start and stop records faster).

The key is always prefixed with 'tablename:entry::'. For example the record in 'subscriber' table for user 'alice@sip.com' has the key: 'subscriber:entry::alice:sip.com'. If all the records are just loaded at startup (or all reloaded at runtime), the key can just be made unique using whatever values added after 'tablename:entry::' prefix. For example, keys for 'address' table records can be: 'address:entry::1', address:entry::2', address:entry::3', ... No 'keys' modparam of 'db_redis' for 'address' table needs to be defined.

Important Note: at this moment the module requires at least one 'keys' parameter, but it does not need to be related to the table loaded from Redis server -- for example, if used only for permissions module with 'address' table, then the 'keys' parameter can be specified for 'version' table.

The mappings can be freely defined in the "keys" module parameter, which is composed of a semi-colon separated list of definitions in the format <table-name>=<entry>:<column-name>[&<map-name>:<column-name>,<column-name>...]. Each table must at least have an "entry" key for db_redis to be able to store data.

Example:

location=entry:ruid&usrdom:username,domain&timer:partition,keepalive;acc=entry:callid,time_hires&cid:callid

For readability purposes, definitions of keys per table can span multiple Kamailio config lines by providing multiple "keys" modparams.

Load the module and set the "db_url" modparam for specific modules to: 'redis://[username]@host:port/database'. Username is optional. The database portion must be a valid Redis database number.

...
loadmodule "db_redis.so"
...
#!define DBURL_USRLOC "redis://127.0.0.1:6379/5"
#!define DBURL_ACC    "redis://127.0.0.1:6379/6"
#!define DBURL_AUTH   "redis://127.0.0.1:6379/7"
#!define DBURL_PERM   "redis://127.0.0.1:6379/8"
...
modparam("db_redis", "schema_path", "/usr/share/kamailio/db_redis/kamailio")
modparam("db_redis", "keys", "version=entry:table_name")
modparam("db_redis", "keys", "location=entry:ruid&usrdom:username,domain&timer:partition,keepalive")
modparam("db_redis", "keys", "acc=entry:callid,time_hires&cid:callid")
modparam("db_redis", "keys", "subscriber=entry:username,domain")
...
modparam("usrloc", "db_url", DBURL_USRLOC)
...
modparam("acc_db", "db_url", DBURL_ACC)
...
modparam("auth_db", "db_url", DBURL_AUTH)
...
modparam("permissions", "db_url", DBURL_PERM)
...

Samples adding records for address table using 'redis-cli':

...
SELECT 8
HMSET address:entry::1 id 1 grp 1 ip_addr "127.0.0.1" mask 32 port 0
HMSET address:entry::2 id 2 grp 1 ip_addr "127.0.0.2" mask 32 port 0
HMSET address:entry::3 id 3 grp 2 ip_addr "127.0.0.3" mask 32 port 0
HMSET address:entry::4 id 4 grp 2 ip_addr "127.0.0.4" mask 32 port 0 tag "test"
...

Note that is some cases, the optional values in database tables can be omitted. For 'address' table, the 'tag' value may be omitted. To avoid any issues, set unused fields to their default values as defined by database schema. When definition allows 'NULL', that field can be unset.