CACHEDB_REDIS module

Admin Guide

Overview

This module is an implementation of a cache system designed to work with a Redis server. It uses hiredis client library to connect to either a single Redis server instance, or to a Redis Server inside a Redis Cluster. It uses the Key-Value interface exported from the core.

Advantages

• memory costs are no longer on the server
• many servers can be used inside a cluster, so the memory is virtually unlimited
• the cache is 100% persistent. A restart of OpenSIPS server will not affect the DB. The Redis DB is also persistent so it can also be restarted without loss of information.
• redis is an open-source project so it can be used to exchange data with various other applications
• By creating a Redis Cluster, multiple OpenSIPS instances can easily share key-value information

Redis Stack Support

Starting with OpenSIPS 3.6, the cachedb_redis module implements the column-oriented cacheDB API functions. This makes it a suitable cacheDB storage in scenarios such as user location federation and full-sharing, which require this API to be available.

The implementation makes use of RedisJSON and RediSearch — these relatively new features are available in Redis Stack Server, instead of the usual Redis Server (Redis OSS project). More documentation is available on the Redis website.

OpenSIPS will auto-detect availability of the RedisJSON support when necessary and log the appropriate messages.

Redis Cluster Support (Topology)

When connecting to a Redis Cluster, the module automatically detects cluster mode and manages the full slot-to-node topology at runtime. No extra configuration is needed beyond the standard cachedb url parameter.

Topology Discovery

At startup, the module probes the Redis server using the CLUSTER SHARDS command (available in Redis 7.0+). If the server does not support this command, it falls back to CLUSTER SLOTS (available in Redis 3.0+). If neither command succeeds, the connection is treated as a single-instance (non-cluster) connection.

The discovered topology is stored internally in an O(1) slot lookup table (16384 slots), mapping each slot directly to its owning master node.

Automatic Topology Refresh

The module automatically refreshes the cluster topology at runtime when any of the following events occur:

• A MOVED redirection is received from a cluster node (indicating a permanent slot migration).
• A connection failure (NULL reply) occurs and the node cannot be reconnected.
• A query targets a slot with no known owner, suggesting the topology is stale.
• An operator triggers a manual refresh via the mi redis cluster refresh MI command.

Automatic refreshes are rate-limited to at most once per second to avoid excessive load on the cluster. The MI-triggered refresh bypasses this rate limit.

MOVED Redirection

The module transparently handles Redis Cluster MOVED redirections:

• MOVED — indicates a permanent slot migration. The module updates its slot map, redirects the query to the new node, and triggers a topology refresh so all future queries go directly to the correct node.

If a redirection points to a node that is not yet known, the module dynamically creates a new node entry, establishes a connection, and retries the query.

Hash Tags

The module supports Redis Cluster hash tags, which allow related keys to be co-located on the same cluster node. If a key contains a {…} substring, only the content between the first { and the next } is used for hash slot calculation. For example, the keys {user1000}.profile and {user1000}.settings will always land on the same node, enabling multi-key operations.

If the braces are empty ({}) or there is no closing brace, the entire key is hashed as usual.

Limitations

• keys (in key:value pairs) may not contain spaces or control characters

Dependencies

OpenSIPS Modules

The following modules must be loaded before this module:

• If a use tls is defined, the tls_mgm and tls_openssl modules will need to be loaded as well.

External Libraries or Applications

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

• hiredis: On the latest Debian based distributions, hiredis can be installed by running ‘apt-get install libhiredis-dev’

Alternatively, if hiredis is not available on your OS repos, hiredis can be downloaded from: https://github.com/antirez/hiredis . Download the archive, extract sources, run make,sudo make install. If TLS connections are enabled via the use tls modparam, hiredis needs to be compiled with TLS support.

Exported Parameters

cachedb_url (string)

The URLs of the server groups that OpenSIPS will connect to in order to use, from script, the cache_store(), cache_fetch(), etc. operations. It may be set more than once. The prefix part of the URL will be the identifier that will be used from the script.

Set cachedb_url parameter

...
# single-instance URLs (Redis Server or Redis Cluster)
modparam("cachedb_redis", "cachedb_url", "redis:group1://localhost:6379/")
modparam("cachedb_redis", "cachedb_url", "redis:cluster1://random_url:8888/")

# multi-instance URL (will perform circular

Use Redis servers

...
cache_store("redis:group1", "key", "$ru value");
cache_fetch("redis:cluster1", "key", $avp(10));
cache_remove("redis:cluster1", "key");
...

Authentication

The module supports three authentication modes based on the URL format:

URL Authentication Formats

URL Format	AUTH Command	Use Case
redis:group://:password@host:port/	AUTH password	Classic Redis (< 6.0) with requirepass
redis:group://username:password@host:port/	AUTH username password	Redis 6+ ACL with per-user credentials
redis:group://host:port/	(none)	Non-authenticated Redis

Caution

For classic password-only authentication, the URL must include a colon before the password (:password@host). Writing password@host without the colon will place the credential in the username field of the URL parser, and authentication will be skipped.

When connecting to a Redis Cluster with authentication, all discovered cluster nodes use the same credentials from the URL.

Unix Socket

Starting with this version, the module supports connecting to a local Redis instance via a Unix domain socket instead of TCP. This can provide lower latency and avoid network overhead for co-located Redis instances.

To use a Unix socket, add a socket= parameter to the URL query string:

# basic Unix socket (no auth)
modparam("cachedb_redis", "cachedb_url",
    "redis:local://localhost/?socket=/var/run/redis/redis.sock")

# Unix socket with password auth
modparam("cachedb_redis", "cachedb_url",
    "redis:local://:password@localhost/?socket=/var/run/redis/redis.sock")

# Unix socket with ACL auth (Redis 6+) and database selection
modparam("cachedb_redis", "cachedb_url",
    "redis:local://user:pass@localhost/2?socket=/var/run/redis/redis.sock")

Constraints:

• Unix socket connections are always treated as single-instance mode (no Redis Cluster support over Unix sockets).
• Unix socket cannot be combined with multiple hosts (failover). Specifying both will cause a startup error.
• TLS is not applicable to Unix socket connections and will be ignored with a warning if use_tls is enabled.
• TCP keepalive is not applicable to Unix sockets and is automatically skipped.

The mi redis cluster info MI command will display Unix socket connections with transport=unix and the socket path. The mi redis ping nodes command works normally with Unix socket connections.

connect_timeout (integer)

This parameter specifies how many milliseconds OpenSIPS should wait for connecting to a Redis node.

Default value is “5000 ms”.

Set connect_timeout parameter

...
# wait 1 second for Redis to connect
modparam("cachedb_redis", "connect_timeout",1000)
...

query_timeout (integer)

This parameter specifies how many milliseconds OpenSIPS should wait for a query response from a Redis node.

Default value is “5000 ms”.

Set query_timeout parameter

...
# wait 1 second for Redis queries
modparam("cachedb_redis", "query_timeout",1000)
...

shutdown_on_error (integer)

By setting this parameter to 1, OpenSIPS will abort startup if the initial connection to Redis is not possible. Runtime reconnect behavior is unaffected by this parameter, and is always enabled.

Default value is “0” (disabled).

Set the shutdown_on_error parameter

...
# abort OpenSIPS startup if Redis is down
modparam("cachedb_redis", "shutdown_on_error", 1)
...

lazy_connect (integer)

By setting this parameter to 1, OpenSIPS will defer establishing Redis connections until the first cache operation is actually performed by each worker process. This prevents idle worker processes (those that never use Redis) from holding open sockets, which avoids sockets getting stuck in CLOSE_WAIT state when Redis is restarted.

When this parameter is enabled, the shutdown on error parameter has no effect, since no connection is attempted at startup time.

Default value is “0” (disabled — connect at startup, preserving existing behavior).

Set the lazy_connect parameter

...
# defer Redis connections until first use
modparam("cachedb_redis", "lazy_connect", 1)
...

use_tls (integer)

Setting this parameter will allow you to use TLS for Redis connections. In order to enable TLS for a specific connection, you can use the “tls_domain=dom_name” URL parameter in the cachedb_url of this module (or other modules that use the CacheDB interface). This should be placed at the end of the URL after the ’?’ character.

When using this parameter, you must also ensure that tls_mgm is loaded and properly configured. Refer to the tls_mgm module for additional info regarding TLS client domains.

Note that TLS is supported by Redis starting with version 6.0. Also, it is an optional feature enabled at compile time and might not be included in the standard Redis packages available for your OS.

Default value is 0 (not enabled)

Set the use_tls parameter

...
modparam("tls_mgm", "client_domain", "redis")
modparam("tls_mgm", "certificate", "[redis]/etc/pki/tls/certs/redis.pem")
modparam("tls_mgm", "private_key", "[redis]/etc/pki/tls/private/redis.key")
modparam("tls_mgm", "ca_list",     "[redis]/etc/pki/tls/certs/ca.pem")
...
modparam("cachedb_redis", "use_tls", 1)
modparam("cachedb_redis", "cachedb_url","redis:tls_group://localhost:6379/?tls_domain=redis")
...

ftsearch_index_name (string)

Only relevant with RedisJSON and RediSearch server-side support.

A global index name to be used for all internal JSON full-text search operations. Future extensions may add, e.g., a connection-level index name setting.

Default value is “idx:usrloc”.

Set the ftsearch_index_name parameter

modparam("cachedb_redis", "ftsearch_index_name", "ix::usrloc")

ftsearch_json_prefix (string)

Only relevant with RedisJSON and RediSearch server-side support.

A key naming prefix for all internally-created Redis JSON objects (e.g. created with JSON.SET or JSON.MSET).

Default value is “usrloc:“.

Set the ftsearch_json_prefix parameter

modparam("cachedb_redis", "ftsearch_json_prefix", "userlocation:")

ftsearch_max_results (integer)

Only relevant with RedisJSON and RediSearch server-side support.

The maximum number of results returned by each internally-triggered FT.SEARCH JSON lookup query.

Default value is 10000 max results.

Set the ftsearch_max_results parameter

modparam("cachedb_redis", "ftsearch_max_results", 100)

redis_keepalive (integer)

TCP keepalive interval in seconds for Redis connections. When set to a positive value, the kernel sends TCP probes on idle connections to detect dead peers (e.g. due to NAT/firewall idle timeout or network partition). This allows the next query to fail immediately instead of waiting for the full query timeout, enabling faster recovery via the existing retry loop.

Set to 0 to disable TCP keepalive. Recommended to keep enabled for production deployments to prevent silent connection death.

Default value is “10” (seconds).

Set redis_keepalive parameter

...
# set TCP keepalive interval to 15 seconds
modparam("cachedb_redis", "redis_keepalive", 15)

# disable TCP keepalive
modparam("cachedb_redis", "redis_keepalive", 0)
...

ftsearch_json_mset_expire (integer)

Only relevant with RedisJSON and RediSearch server-side support.

A Redis EXPIRE timer to set/refresh on the JSON key after each JSON.MSET operation (create the JSON or add/remove subkeys), in seconds. A value of 0 disables the EXPIRE queries completely.

Default value is 3600 seconds.

Set the ftsearch_json_mset_expire parameter

modparam("cachedb_redis", "ftsearch_json_mset_expire", 7200)

Exported Functions

The module does not export functions to be used in configuration script.

Exported MI Functions

redis_cluster_info

Displays detailed information about all Redis connections managed by the module, including cluster topology, per-node connection status, slot assignments, and per-node query counters.

Parameters:

• group (optional) - if specified, only connections belonging to this group will be listed (e.g. “local” from a “redis:local://…” URL). If omitted, all Redis connections are listed.

The response is a JSON array of connection objects. Each connection object includes:

• group - the connection group name
• url - the original cachedb_url
• mode - “cluster” or “single”
• cluster_command (cluster mode only) - “SHARDS” or “SLOTS”, depending on which Redis command is used for topology discovery
• topology_refreshes - number of topology refreshes performed on this connection
• last_topology_refresh - UNIX timestamp of the last topology refresh
• nodes - array of cluster node objects, each containing: ip, port, status (“connected”/“disconnected”), slots_assigned (cluster mode only), queries, errors, moved, last_activity (seconds since last successful query, -1 if never queried)
• total_slots_mapped (cluster mode only) - total number of slots with an assigned node (should be 16384 for a healthy cluster)

MI FIFO Command Format:

## list all Redis connections
opensips-cli -x mi redis_cluster_info

## list only the "local" group
opensips-cli -x mi redis_cluster_info group=local

redis_cluster_refresh

Forces an immediate topology refresh on Redis Cluster connections. This bypasses the normal once-per-second rate limit and queries the cluster for its current slot-to-node mapping. Useful after manual cluster rebalancing or node additions/removals.

For non-cluster (single instance) connections, the command returns a “skipped (not cluster mode)” status.

Parameters:

• group (optional) - if specified, only the connection belonging to this group will be refreshed. If omitted, all cluster connections are refreshed.

The response is a JSON array of objects, one per connection, each containing group and status (“ok”, “error”, or “skipped (not cluster mode)”).

MI FIFO Command Format:

## refresh all cluster connections
opensips-cli -x mi redis_cluster_refresh

## refresh only the "local" group
opensips-cli -x mi redis_cluster_refresh group=local

redis_ping_nodes

Sends a PING command to each Redis node and reports per-node reachability status with round-trip latency. Useful for on-demand health checks without waiting for the next query.

Parameters:

• group (optional) - if specified, only nodes belonging to this group will be pinged. If omitted, all Redis connections are pinged.

The response is a JSON array of connection objects. Each connection object includes:

• group - the connection group name
• nodes - array of node objects, each containing: ip, port, status (“reachable”, “unreachable”, or “disconnected”), latency_us (round-trip time in microseconds, -1 if not reachable)

MI FIFO Command Format:

## ping all Redis nodes
opensips-cli -x mi redis_ping_nodes

## ping only the "local" group
opensips-cli -x mi redis_ping_nodes group=local

Exported Statistics

redis_queries

Total number of successful Redis queries executed across all connections and processes.

redis_queries_failed

Total number of failed Redis queries (NULL replies from hiredis or Redis error responses other than MOVED).

redis_moved

Total number of MOVED redirections received from Redis Cluster nodes. A MOVED response indicates a permanent slot migration - the module updates its slot map and retries the query on the correct node.

redis_topology_refreshes

Total number of cluster topology refreshes performed (via CLUSTER SHARDS or CLUSTER SLOTS). This counter increments both for automatic refreshes (triggered by MOVED responses or unreachable nodes) and manual refreshes (triggered via the mi redis cluster refresh MI command).

Raw Query Syntax

The cachedb_redis module allows to run RAW queries, thus taking full advantage of the capabilities of the back-end.

The query syntax is the typical REDIS one.

Here are a couple examples of running some Redis queries :

Redis Raw Query Examples

...
  $var(my_hash) = "my_hash_name";
  $var(my_key) = "my_key_name";
  $var(my_value) = "my_key_value";
  cache_raw_query("redis","HSET $var(my_hash) $var(my_key) $var(my_value)");
  cache_raw_query("redis","HGET $var(my_hash) $var(my_key)","$avp(result)");
  xlog("We have fetched $avp(result) \n");
...
  $var(my_hash) = "my_hash_name";
  $var(my_key1) = "my_key1_name";
  $var(my_key2) = "my_key2_name";
  $var(my_value1) = "my_key1_value";
  $var(my_value2) = "my_key2_value";
  cache_raw_query("redis","HSET $var(my_hash) $var(my_key1) $var(my_value1)");
  cache_raw_query("redis","HSET $var(my_hash) $var(my_key2) $var(my_value2)");
  cache_raw_query("redis","HGETALL $var(my_hash)","$avp(result)");

  $var(it) = 0;
  while ($(avp(result)[$var(it)]) != NULL) {
    xlog("Multiple key reply: - we have fetched $(avp(result)[$var(it)]) \n");
    $var(it) = $var(it) + 1;
  }
...

Frequently Asked Questions

Q: My OpenSIPS is occasionally crashing in libhiredis, what to do?

Make sure you’ve upgraded the Redis “libhiredis” client library to at least version 0.14.1. There was at least one significant vulnerability reported in library versions prior to that one (CVE-2020-7105), so upgrading to latest stable may very well fix the crash!

Contributors

By Commit Statistics

Top contributors by DevScore⁽¹⁾, authored commits⁽²⁾ and lines added/removed⁽³⁾

#	Name	DevScore	Commits	Lines++	Lines—
1.	Debian	65	3	6339	428
2.	Liviu Chircu (@liviuchircu)	44	23	1759	327
3.	Vlad Paiu (@vladpaiu)	33	19	1451	55
4.	Razvan Crainea (@razvancrainea)	21	17	205	97
5.	Stefan Darius	17	5	950	193
6.	Vlad Patrascu (@rvlad-patrascu)	13	6	603	46
7.	Norman Brandinger (@NormB)	10	6	267	11
8.	Bogdan-Andrei Iancu (@bogdan-iancu)	7	5	10	8
9.	Maksym Sobolyev (@sobomax)	6	4	7	7
10.	jalung	5	1	168	85

All remaining contributors: rdondeti, Dan Pascu (@danpascu), Kristian Høgh, Ezequiel Lovelle (@lovelle), John Burke (@john08burke), tcresson, zhengsh, Ken Rice, Eddie Fiorentine, Peter Lemenkov (@lemenkov), Julián Moreno Patiño, Jarrod Baumann (@jarrodb).

(1) DevScore = author_commits + author_lines_added / (project_lines_added / project_commits) + author_lines_deleted / (project_lines_deleted / project_commits)

(2) including any documentation-related commits, excluding merge commits

(3) ignoring whitespace edits, renamed files and auto-generated files

By Commit Activity

#	Name	Commit Activity
1.	Stefan Darius	Jun 2026 - Jul 2026
2.	Razvan Crainea (@razvancrainea)	Feb 2012 - Jun 2026
3.	Norman Brandinger (@NormB)	May 2025 - Apr 2026
4.	Debian	Mar 2026 - Mar 2026
5.	rdondeti	Mar 2026 - Mar 2026
6.	Ken Rice	Sep 2025 - Sep 2025
7.	Eddie Fiorentine	Mar 2025 - Mar 2025
8.	Liviu Chircu (@liviuchircu)	Mar 2014 - Jan 2025
9.	Vlad Paiu (@vladpaiu)	Oct 2011 - Nov 2024
10.	tcresson	Oct 2023 - Oct 2023

All remaining contributors: zhengsh, Maksym Sobolyev (@sobomax), John Burke (@john08burke), Vlad Patrascu (@rvlad-patrascu), Bogdan-Andrei Iancu (@bogdan-iancu), Dan Pascu (@danpascu), Peter Lemenkov (@lemenkov), Kristian Høgh, Julián Moreno Patiño, Jarrod Baumann (@jarrodb), jalung, Ezequiel Lovelle (@lovelle).

(1) including any documentation-related commits, excluding merge commits

Documentation

Contributors

Last edited by: Razvan Crainea (@razvancrainea), Debian, Liviu Chircu (@liviuchircu), Vlad Patrascu (@rvlad-patrascu), Peter Lemenkov (@lemenkov), Julián Moreno Patiño, Vlad Paiu (@vladpaiu).

License

All documentation files (i.e. .md extension) are licensed under the Creative Common License 4.0