CLUSTERER Module


Table of Contents

1. Admin Guide
1.1. Overview
1.2. Capabilities layer
1.3. Dependencies
1.3.1. OpenSIPS Modules
1.3.2. External Libraries or Applications
1.4. Exported Parameters
1.4.1. my_node_id
1.4.2. db_mode
1.4.3. db_url
1.4.4. db_table
1.4.5. sharing_tag
1.4.6. my_node_info
1.4.7. neighbor_node_info
1.4.8. ping_interval
1.4.9. ping_timeout
1.4.10. node_timeout
1.4.11. seed_fallback_interval
1.4.12. sync_packet_max_size
1.4.13. id_col
1.4.14. cluster_id_col
1.4.15. node_id_col
1.4.16. url_col
1.4.17. state_col
1.4.18. no_ping_retries_col
1.4.19. priority_col
1.4.20. sip_addr_col
1.4.21. flags_col
1.4.22. description_col
1.5. Exported Functions
1.5.1. cluster_send_req(cluster_id, dst_id, msg, [tag])
1.5.2. cluster_send_rpl(cluster_id, dst_id, msg, tag)
1.5.3. cluster_broadcast_req(cluster_id, msg, [tag])
1.5.4. cluster_check_addr(cluster_id, ip, addr_type)
1.6. Exported MI Functions
1.6.1. clusterer_reload
1.6.2. clusterer_list
1.6.3. clusterer_list_topology
1.6.4. clusterer_set_status
1.6.5. cluster_send_mi
1.6.6. cluster_broadcast_mi
1.6.7. clusterer_list_cap
1.6.8. clusterer_shtag_set_active
1.6.9. clusterer_list_shtags
1.7. Exported Script Variables
1.7.1. $cluster.sh_tag
1.8. Exported Events
1.8.1. E_CLUSTERER_REQ_RECEIVED
1.8.2. E_CLUSTERER_RPL_RECEIVED
1.8.3. E_CLUSTERER_NODE_STATE_CHANGED
1.9. Usage Example
2. Developer Guide
2.1. Available Functions
2.1.1. get_nodes(cluster_id)
2.1.2. free_nodes(list)
2.1.3. set_state(cluster_id, state)
2.1.4. check_addr(cluster_id, su)
2.1.5. get_my_id()
2.1.6. send_to(packet, cluster_id, node_id)
2.1.7. send_all(packet, cluster_id)
2.1.8. get_next_hop(cluster_id, node_id)
2.1.9. free_next_hop(next_hop)
2.1.10. register_module(mod_name, cb, auth_check, accept_clusters_ids, no_accept_clusters)
3. Contributors
3.1. By Commit Statistics
3.2. By Commit Activity
4. Documentation
4.1. Contributors

List of Tables

3.1. Top contributors by DevScore(1), authored commits(2) and lines added/removed(3)
3.2. Most recently active contributors(1) to this module

List of Examples

1.1. Set my_node_id parameter
1.2. Set db_mode parameter
1.3. Set db_url parameter
1.4. Set db_table parameter
1.5. Set sharing_tag parameter
1.6. Set my_node_info parameter
1.7. Set neighbor_node_info parameter
1.8. Set ping_interval parameter
1.9. Set ping_timeout parameter
1.10. Set node_timeout parameter
1.11. Set seed_fallback_interval parameter
1.12. Set sync_packet_max_size parameter
1.13. Set id_col parameter
1.14. Set cluster_id_col parameter
1.15. Set node_id_col parameter
1.16. Set url_col parameter
1.17. Set state_col parameter
1.18. Set no_ping_retries_col parameter
1.19. Set priority_col parameter
1.20. Set sip_addr_col parameter
1.21. Set flags_col parameter
1.22. Set description_col parameter
1.23. cluster_send_req() usage
1.24. cluster_send_rpl() usage
1.25. cluster_broadcast_req() usage
1.26. cluster_check_addr() usage
1.27. clusterer_list usage
1.28. clusterer_list_topology usage
1.29. clusterer_list_cap usage
1.30. Example database content - clusterer table
1.31. Node A configuration
1.32. Node B configuration

Chapter 1. Admin Guide

1.1. Overview

The clusterer module is used to organize multiple OpenSIPS instances into groups(clusters) in which the nodes can communicate with each other in order to replicate, share information or perform distributed tasks. The distributed logic is performed either by different modules that use the clusterer interface (i.e. the dialog module can replicate dialogs/profiles, the ratelimit module can share pipes across multiple instances etc.) or at the script level. The clusterer module itself only provides an interface to send/receive BIN packets and get notifications about node availability. It achieves this by internally learning the cluster topology and state of the nodes. Provisioning the nodes within a cluster is done over the database or through the configuration script. The node-related information can be checked and triggered to be reloaded by sending commands over the MI interface.

The topology established by the clusterer module is an overlay of nodes where the "links" represent communication availability at BIN interface level. For this purpose, a probing mechanism is used, consisting of regular pings to all nodes in a cluster for which replies must be received within a given interval. All nodes in the cluster exchange information about the state of their links with other nodes and compute a "routing table" which gives a next hop for each destination. The metric for the shortest path is the number of hops. When there is no direct link to a destination, the BIN packet sent by a module is transparently routed through the cluster.

Note that an OpenSIPS instance can belong to multiple clusters, communicating and establishing the topology separately for each one. In order to provision this in the database or the script, each node has an unique ID at global level, which can be referenced in each cluster.

An OpenSIPS instance can dynamically learn all the nodes in the cluster that are not provisioned in the database. For instances that do not load any node info from the DB, it is enough to define at least one neigbour in the script in order to discover all the cluster components.

1.2. Capabilities layer

The clusterer module also keeps track of the state of the nodes in terms of data synchronization for the functionalities (or "capabilities") implemented on top by other modules. Some capabilities require a full data sync(at OpenSIPS startup or at runtime via MI) from a valid "donor" node in the cluster that has the full data set. Furthermore, a capability can query the clusterer module in order to partition some distributed logic only over the synchronized nodes in the cluster.

Each node in the cluster starts with an empty dataset and tries to find a suitable node to pull data from. In order to help "bootstrap" the cluster, a "seed" node should be defined. This is done by setting the value seed for the flags column in the clusterer table(or the property with the same name in the my_node_info parameter). The seed node will simply fall back to a "synced" state after a configurable interval( seed_fallback_interval parameter). This interval can also be set to 0, in which case the seed node will be considered already synced from startup. Note that configuring a a seed node is required only for capabilities that synchronize data at startup, so check the corresponding modules documentation.

The clusterer module transparently exposes the sip_addr column from the clusterer table(or the property with the same name in the my_node_info parameter) to the modules on top so check the corresponding modules documentation for the use of this node related information.

1.3. Dependencies

1.3.1. OpenSIPS Modules

The following modules must be loaded before this module:

  • a database module - if db_mode is 1.

  • proto_bin module.

1.3.2. External Libraries or Applications

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

  • None.

1.4. Exported Parameters

1.4.1. my_node_id

The id of the local instance. This parameter must be equal to one of the node_id fields in the database.

No default value. This parameter must be explicitly set to a value greater than zero.

Example 1.1. Set my_node_id parameter

...
modparam("clusterer", "my_node_id", 1)
...
		

1.4.2. db_mode

Specifies whether the node information for the local instance, as well as other instances in the cluster, should be loaded from the database or configured in the script(see my_node_info and neighbor_node_info). A value of 0 means that DB is not used and the cluster topology in terms of node information will be discovered dynamically at runtime.

Default value is 1

Example 1.2. Set db_mode parameter

...
modparam("clusterer", "db_mode", 0)
...
		

1.4.3. db_url

The database url.

Default value is NULL.

Example 1.3. Set db_url parameter

...
modparam("clusterer", "db_url",
	"mysql://opensips:opensipsrw@localhost/opensips")
...
		

1.4.4. db_table

The name of the table storing the clustering information.

Default value is clusterer.

Example 1.4. Set db_table parameter

...
modparam("clusterer", "db_table", "clusterer")
...
		

1.4.5. sharing_tag

The definition of a sharing tag. The sharing tag is managed by the clusterer module, but can be used (in terms of reading its state) by any module build on top of clusterer engine, like dialog or presence.

Note that ohter tags may be dynamically learned during runtime via clustering communication with other nodes.

The format for this value is tag_name / cluster_id = active/backup.

Multiple definitions of this parameter are allowed. The default value is none.

Example 1.5. Set sharing_tag parameter

...
modparam("clusterer", "sharing_tag", "vip1/2=active")
modparam("clusterer", "sharing_tag", "node/10=backup")
...

1.4.6. my_node_info

Node specification similar to the information provided by a row in the clusterer DB table corresponding to the local instance. This parameter can be set multiple times in order to include the local node in multiple clusters.

Parameter format: multiple "prop=value" property definitions separated by ',' where the name of the properties is the same as the DB column names. At least the cluster_id and url properties must be defined.

This parameter is required if db_mode is set to 0 in order to properly advertise information about the local instance in the dynamic node learning process.

Example 1.6. Set my_node_info parameter

...
modparam("clusterer", "my_node_info", "cluster_id=1, url=bin:192.168.0.5:5566")
...
		

1.4.7. neighbor_node_info

Node specification similar to the information provided by a row in the clusterer DB table corresponding to another instance in the cluster. This node will be the entry point in the cluster for the local instance in the dynamic node learning process. This parameter can be set multiple times to define multiple neigbors to connect to (or the same neighbor but in different clusters).

Parameter format: multiple "prop=value" property definitions separated by ',' where the name of the properties is the same as the DB column names. At least the cluster_id, node_id and url properties must be defined.

This parameter should be set at least once if db_mode is set to 0 in order to properly learn the cluster topology. If not set, the only way to learn the node topology is by other nodes connecting to the local instance.

Example 1.7. Set neighbor_node_info parameter

...
modparam("clusterer", "neighbor_node_info", "cluster_id=1,node_id=2,url=bin:192.168.0.6:5566")
...
		

1.4.8. ping_interval

The interval in seconds between regular pings sent to a neighbour node.

Default value is 4

Example 1.8. Set ping_interval parameter

...
modparam("clusterer", "ping_interval", 1)
...
		

1.4.9. ping_timeout

The time in milliseconds to wait for a reply to a previously sent ping before retrying or considering the link with the neighbour node down. This is also the interval between successive retries if the send fails.

Default value is 1000

Example 1.9. Set ping_timeout parameter

...
modparam("clusterer", "ping_timeout", 500)
...
		

1.4.10. node_timeout

The time in seconds to wait before pinging is restarted for a failed node.

Default value is 60

Example 1.10. Set node_timeout parameter

...
modparam("clusterer", "node_timeout", 10)
...
		

1.4.11. seed_fallback_interval

Only relevant for "seed" nodes. The time, in seconds, to wait for a suitable donor node before falling back to a "synced" state, following a node restart or an MI cluster sync command. If set to 0, the node is considered as already synced from startup.

Default value is 0.

Example 1.11. Set seed_fallback_interval parameter

...
modparam("clusterer", "seed_fallback_interval", 10)
...
		

1.4.12. sync_packet_max_size

The maximum size of the BIN packets sent while doing data synchronization. This is only a suggested value as the actual size of the packets may be slightly larger.

Default value is 65535.

Example 1.12. Set sync_packet_max_size parameter

...
modparam("clusterer", "sync_packet_max_size", 32765)
...
		

1.4.13. id_col

The name of the column storing an id for the table rows.

Default value is id.

Example 1.13. Set id_col parameter

...
modparam("clusterer", "id_col", "id")
...
		

1.4.14. cluster_id_col

The name of the column to store the id of a cluster.

Default value is cluster_id.

Example 1.14. Set cluster_id_col parameter

...
modparam("clusterer", "cluster_id_col", "cluster_id")
...
		

1.4.15. node_id_col

The name of the column to store the id of an instance. The values must be greater than 0.

Default value is node_id.

Example 1.15. Set node_id_col parameter

...
modparam("clusterer", "node_id_col", "node_id")
...
		

1.4.16. url_col

The name of the column containing the instance url. The values must be greater than 0.

Default value is url.

Example 1.16. Set url_col parameter

...
modparam("clusterer", "url_col", "url")
...
		

1.4.17. state_col

The name of the column storing the state of the node(enabled/disabled).

Default value is state.

Example 1.17. Set state_col parameter

...
modparam("clusterer", "state_col", "state")
...
		

1.4.18. no_ping_retries_col

The name of the column containing the maximum number of ping retries before the link with the neighbour node is considered down.

Default value is no_ping_retries.

Example 1.18. Set no_ping_retries_col parameter

...
modparam("clusterer", "no_ping_retries_col", "no_ping_retries")
...
		

1.4.19. priority_col

The name of the column storing the node priority to be chosen as next hop in case of same length(number of hops) paths when rerouting messages.

Default value is priority.

Example 1.19. Set priority_col parameter

...
modparam("clusterer", "priority_col", "priority")
...
		

1.4.20. sip_addr_col

The name of the column containing a SIP address for the node.

Default value is sip_addr.

Example 1.20. Set sip_addr_col parameter

...
modparam("clusterer", "sip_addr_col", "sip_addr")
...
		

1.4.21. flags_col

The name of the column containing the node flags.

Default value is flags.

Example 1.21. Set flags_col parameter

...
modparam("clusterer", "flags_col", "flags")
...
		

1.4.22. description_col

The name of the column containing a node description.

Default value is description.

Example 1.22. Set description_col parameter

...
modparam("clusterer", "description_col", "description")
...
		

1.5. Exported Functions

1.5.1.  cluster_send_req(cluster_id, dst_id, msg, [tag])

This function is used to send a generic, request-like message, containing custom data, to a specific node in a cluster, directly from the script. The message is not a "request" per se but according to the logic on the receiving side, that node can send back a reply. In order to correlate a received reply with the request sent out, the function returns, through the tag parameter, a randomly generated communication tag, which is sent along in the the original message, that can be checked against the tag received in a reply.

Meaning of the parameters is as follows:

  • cluster_id (int) - the cluster ID of the destination node;

  • dst_id (int) - the ID of the destiantion node;

  • msg (string) - actual message payload;

  • tag (var, optional) - randomly generated communication tag.

The function can return the following values:

  • 1 - successfully sent message to destination node or a valid next hop

  • -1 - local node is disabled so sending is impossbile

  • -2 - destination node is not reachable through any path according to the discovered topology

  • -3 - destination node or valid next hop appear to be reachable but send failed or other OpenSIPS internal error

This function can be used from REQUEST_ROUTE, FAILURE_ROUTE, ONREPLY_ROUTE, BRANCH_ROUTE, LOCAL_ROUTE and EVENT_ROUTE.

Example 1.23. cluster_send_req() usage

...
# send a request
cluster_send_req(1, 1, "Check USER: $fU", $var(req_tag));
# wait for reply
$avp(filter) = "tag=" + $var(req_tag);
async(wait_for_event("E_CLUSTERER_RPL_RECEIVED", $avp(filter), 5), rpl_resume);
# done
...
route[rpl_resume] {
  xlog("Received reply: $avp(msg)\n");
}
...
				

1.5.2.  cluster_send_rpl(cluster_id, dst_id, msg, tag)

This function is used to send a generic, reply-like message, containing custom data, to a specific node in a cluster, directly from the script. The message is marked as a "reply" so this function should ony be used for replying to a previously request-like message received. In order for the other node, which initially sent a request, to be able to correlate it with this reply, a communication tag, received along with the request, should be passed to the function.

Meaning of the parameters is as follows:

  • cluster_id (int) - the cluster ID of the destination node;

  • dst_id (int) - the ID of the destiantion node;

  • msg (string) - actual message payload;

  • tag (var) - communication tag.

The function can return the following values:

  • 1 - successfully sent message to destination node or a valid next hop

  • -1 - local node is disabled so sending is impossbile

  • -2 - destination node is not reachable through any path according to the discovered topology

  • -3 - destination node or valid next hop appear to be reachable but send failed or other OpenSIPS internal error

This function can be used from REQUEST_ROUTE, FAILURE_ROUTE, ONREPLY_ROUTE, BRANCH_ROUTE, LOCAL_ROUTE and EVENT_ROUTE.

Example 1.24. cluster_send_rpl() usage

...
event_route[E_CLUSTERER_REQ_RECEIVED] {
  cluster_send_rpl($param(cluster_id), $param(src_id), $var(my_reply), $param(tag));
}
...
				

1.5.3.  cluster_broadcast_req(cluster_id, msg, [tag])

This function has a similar behaviour to the cluster_send_req() function with the exception that the message is sent to all the nodes in the specified cluster.

The function can return the following values:

  • 1 - successfully sent message to at least one node;

  • -1 - local node is disabled so sending is impossbile;

  • -2 - all nodes in the cluster are unreachable according to the discovered topology;

  • -3 - send failed for all nodes in the cluster or other OpenSIPS internal error.

The meaning of the parameters is the same as for cluster_send_req().

This function can be used from REQUEST_ROUTE, FAILURE_ROUTE, ONREPLY_ROUTE, BRANCH_ROUTE, LOCAL_ROUTE and EVENT_ROUTE.

Example 1.25. cluster_broadcast_req() usage

...
cluster_broadcast_req($var(cl_id), $var(share_data));
...
				

1.5.4.  cluster_check_addr(cluster_id, ip, addr_type)

This function checks whether the given IP address belongs to one of the nodes in the cluster.

Parameters:

  • cluster_id (int)

  • ip (string)

  • addr_type (string, optional) - select the address of the node that the comparison is made against, with the possible values of:

    • bin - node's BIN interface listener;

    • sip - node's DB provisioned SIP address.

    The default behaviour is to compare against the SIP address.

This function can be used from REQUEST_ROUTE, FAILURE_ROUTE, ONREPLY_ROUTE, BRANCH_ROUTE, LOCAL_ROUTE and EVENT_ROUTE.

Example 1.26. cluster_check_addr() usage

...
if (cluster_check_addr(1, $si)) {
	...
}
...
				

1.6. Exported MI Functions

1.6.1.  clusterer_reload

Reloads data from the clusterer database. The currently established topology will be lost and the node will rediscover the new topology.

Name: clusterer_reload

Parameters:none

MI FIFO Command Format:

		opensips-cli -x mi clusterer_reload
		

1.6.2.  clusterer_list

Lists information(node id, URL, link state with that node etc.) about the other nodes in each cluster.

Name: clusterer_list

Parameters:none

Example 1.27. clusterer_list usage

$ opensips-cli -x mi clusterer_list
{
    "Clusters": [
        {
            "cluster_id": 1,
            "Nodes": [
                {
                    "node_id": 1,
                    "db_id": 1,
                    "url": "bin:127.0.0.1",
                    "Enabled": true,
                    "link_state": "Up     ",
                    "next_hop": "1",
                    "description": "none"
                }
            ]
        }
    ]
}

1.6.3.  clusterer_list_topology

Lists each cluster's topology from the local node's perspective as an adjacency list. A node appears as a neighbour if the link with that node is up.

Note that if a node id appears in multiple clusters, it refers to the same instance that belongs to different clusters, for which it has a different topology.

Name: clusterer_list_topology

Parameters:none

Example 1.28. clusterer_list_topology usage

$ opensips-cli -x mi clusterer_list_topology
{
    "Clusters": [
        {
            "cluster_id": 1,
            "Nodes": [
                {
                    "node_id": 2,
                    "Neighbours": [
                        1
                    ]
                },
                {
                    "node_id": 1,
                    "Neighbours": [
                        2
                    ]
                }
            ]
        }
    ]
}

1.6.4.  clusterer_set_status

Sets the status(Enabled/Disabled) of the local node in a specified cluster. A disabled node does not send any messages and ignores received ones thus appearing as a failed node in the topology.

Name: clusterer_set_status

Parameters:

  • cluster_id - indicates the id of the cluster.

  • status - indicates the new status(0 - Disabled, 1 - Enabled).

MI FIFO Command Format:

		opensips-cli -x mi clusterer_set_status 1 0
		

1.6.5.  cluster_send_mi

Dispatches a given MI command to be run on a specific node in the cluster.

Name: cluster_send_mi

Parameters:

  • cluster_id - id of the cluster.

  • destination - id of the destination node

  • cmd_name - name of the MI command to be run

  • cmd_params (optional) - array of parameters for the MI command to be run

Note that MI commands that require named parameters or arrays as parameter values are not currently supported.

MI FIFO Command Format:

opensips-cli -x mi cluster_send_mi 1 3 lb_reload
		

1.6.6.  cluster_broadcast_mi

Dispatches a given MI command to be run on all the nodes in a cluster. The command is also executed locally.

Name: cluster_broadcast_mi

Parameters:

  • cluster_id - id of the cluster.

  • cmd_name - name of the MI command to be run

  • cmd_params (optional) - array of parameters for the MI command to be run

Note that MI commands that require named parameters or arrays as parameter values are not currently supported.

MI FIFO Command Format:

opensips-cli -x mi cluster_broadcast_mi 1 dr_reload partition_5
		

1.6.7.  clusterer_list_cap

Lists the registered capabilities and their states.

Name: clusterer_list_cap

Parameters:none

Example 1.29. clusterer_list_cap usage

$ opensips-cli -x mi clusterer_list_cap
{
    "Clusters": [
        {
            "cluster_id": 1,
            "Capabilities": [
                {
                    "name": "dialog-dlg-repl",
                    "state": "Ok"
                },
                {
                    "name": "dialog-prof-repl",
                    "state": "Ok"
                }
            ]
        }
    ]
}

1.6.8. clusterer_shtag_set_active

Set the given sharing tag to the active state. The information about this change is also broadcasted in the cluster in order to force any other node that may be active on this tag to step down to backup.

Name: clusterer_shtag_set_active

Parameters: tag - the name of the tag to be set active and the cluster it belogs to, in the format 'tag/cluster_id'.

MI FIFO Command Format:

		opensips-cli -x mi clusterer_shtag_set_active vip1/3
		

1.6.9. clusterer_list_shtags

Lists all known sharing tags and their states.

Name: clusterer_list_shtags

Parameters: Command takes no parameters

MI FIFO Command Format:

		opensips-cli -x mi clusterer_list_shtags
		

1.7. Exported Script Variables

1.7.1. $cluster.sh_tag

This is a read/write variable that allows access to the sharing tags managed by the clusterer module.

The name of such a variable has the format of tag_name/cluster_id, like $cluster.sh_tag(vip/3) accessing the sharing tag "vip" from cluster ID 3.

When setting, a sharing tag may be only switched to active by assigned it:

  • "active" string value
  • 1 or higher numerical value

When reading it value, a sharing tag returns:

  • "active" or 1 if active
  • "backup" or 0 if backup

A NULL value may returned only as a result of an internal error (like memory errors).

1.8. Exported Events

1.8.1.  E_CLUSTERER_REQ_RECEIVED

This event is raised when a generic, request-like, clusterer message is received. This type of message is sent directly from the script and not by an OpenSIPS module.

Parameters:

  • cluster_id - The cluster ID of the source node.

  • src_id - The ID of the source node.

  • msg - The actual message payload.

  • tag - The communication tag of this message, generated by the source node. This could be used to send a reply corresponding to the received message by providing the tag to the cluster_send_rpl() function.

1.8.2.  E_CLUSTERER_RPL_RECEIVED

This event is raised when a generic, reply-like, clusterer message is received. This type of message is sent directly from the script and not by an OpenSIPS module.

Parameters:

  • cluster_id - The cluster ID of the source node.

  • src_id - The ID of the source node.

  • msg - The actual message payload.

  • tag - The communication tag of this message. This could be used to match the received reply with a request sent with the cluster_send_req() or cluster_broadcast_req() functions.

1.8.3.  E_CLUSTERER_NODE_STATE_CHANGED

This event is raised when the state of a node changes in terms of availability.

Parameters:

  • cluster_id - The cluster ID.

  • node_id - The ID of the node.

  • new_state - The new state of the node, with the possible values: 0 - down, 1 - up.

1.9. Usage Example

This section provides an usage example for replicating ratelimit pipes between two OpenSIPS instances. It uses the clusterer module to manage the replicating nodes, and along with the proto_bin module, to send the replicated information.

The setup topology is simple: we have two OpenSIPS nodes running on two separate machines (although they could run on the same machine as well): Node A has IP 192.168.0.5 and Node B has IP 192.168.0.6. Both have, besides the traffic listeners (UDP, TCP, etc.), BIN listeners bound on port 5566. These listeners will be used for the binary communication.

We insert in the the clusterer table the following:

Example 1.30. Example database content - clusterer table

+----+------------+---------+----------------------+-------+-----------------+----------+----------+-------+-------------+
| id | cluster_id | node_id | url                  | state | no_ping_retries | priority | sip_addr | flags | description |
+----+------------+---------+----------------------+-------+-----------------+----------+----------+-------+-------------+
| 10 |          1 |       1 | bin:192.168.0.5:5566 |     1 |                3|       50 | NULL     | NULL  | Node A      |
| 20 |          1 |       2 | bin:192.168.0.6:5566 |     1 |                3|       50 | NULL     | NULL  | Node B      |
+----+------------+---------+----------------------+-------+-----------------+----------+----------+-------+-------------+
		

  • cluster_id - identifier of the cluster. All nodes within a group/cluster should have the same id (in our example, both nodes have ID 1). The values must be greater than 0.

  • node_id - identifier of the machine/node so each instance within a cluster should have a different ID. The values must be greater than 0. In our example, Node A will have ID 1, and Node B ID 2.

  • url - address where all the BIN packets for that instance will be sent to.

  • state - state of the node: 1 means Enabled, 0 means Disabled. A disabled node will not send any BIN packets and will drop received ones.

  • no_ping_retries - maximum number of ping retries before the link with a node is considered down.

  • priority - the priority of a node to be chosen as next hop in case of same length(number of hops) paths when rerouting messages; it is not relevant for this two-node topology example.

  • sip_addr - SIP address for the node that is transparently provided to modules; it has no use for the ratelimit module in our example.

  • flags - used to define a seed node; it has no use in our example.

  • description - an opaque value used to describe the node

After provisioning the two nodes in the database, we have to configure the two instances of OpenSIPS. First, we configure Node A:

Example 1.31. Node A configuration

...
listen = bin:192.168.0.5:5566 # bin listener for Node A

loadmodule "proto_bin.so"

loadmodule "clusterer.so"
modparam("clusterer", "db_url", "mysql://opensips@192.168.0.7/opensips")
modparam("clusterer", "my_node_id", 1) # node_id for Node A

loadmodule "ratelimit.so"
modparam("ratelimit", "pipe_replication_cluster", 1)
...
		

Similarly, the configuration for Node B is as follows:

Example 1.32. Node B configuration

...
listen = bin:192.168.0.6:5566 # bin listener for Node B

loadmodule "proto_bin.so"

loadmodule "clusterer.so"
# ideally, use the same database for both nodes
modparam("clusterer", "db_url", "mysql://opensips@192.168.0.7/opensips")
modparam("clusterer", "my_node_id", 2) # node_id for Node B

loadmodule "ratelimit.so"
modparam("ratelimit", "pipe_replication_cluster", 1)
...
		

Starting the two OpenSIPS instances with the above configurations provides your platform the ability to used shared ratelimit pipes in a very efficient and scalable way.

Chapter 2. Developer Guide

2.1. Available Functions

2.1.1.  get_nodes(cluster_id)

This function will return a list of all the reachable nodes(if the direct link is down/probing, a path through intermediary nodes is considered) in the specified cluster.

The returned nodes structure:

...
typedef struct clusterer_node {
    int node_id;
    union sockaddr_union addr;
    str sip_addr;
    str description;
    struct clusterer_node *next;
} clusterer_node_t;
...
        

Meaning of the parameters is as follows:

  • int cluster_id - the cluster id

2.1.2.  free_nodes(list)

This function will free the lits of nodes returned by get_nodes.

Meaning of the parameters is as follows:

  • clusterer_node_t *list - list header

2.1.3.  set_state(cluster_id, state)

This function sets the state(enabled/disabled) of the current node in the specified cluster.

Meaning of the parameters is as follows:

  • int cluster_id - the cluster id

  • enum cl_node_state state - the new state; possible values:

    • STATE_DISABLED

    • STATE_ENABLED

2.1.4.  check_addr(cluster_id, su)

This function checks if a given address belongs to one of the nodes in the cluster.

Meaning of the parameters is as follows:

  • int cluster_id - the cluster id

  • union sockaddr_union* su - socket address

2.1.5.  get_my_id()

This function will return the id of the current node.

2.1.6.  send_to(packet, cluster_id, node_id)

This functon will send the given BIN packet to the specified node in the cluster. If the direct link is down/probing, it will send the packet to an intermediary node if the destination node is reachable through another path in the cluster topology.

Meaning of the parameters is as follows:

  • bin_packet_t packet - the packet to be sent

  • int cluster_id - the cluster id

  • int node_id - the id of the destination node

The function returns one of the following:

  • CLUSTERER_SEND_SUCCESS - successfully sent packet to destination node or a valid next hop

  • CLUSTERER_CURR_DISABLED - current node is disabled so sending is impossbile

  • CLUSTERER_DEST_DOWN - destination node is not reachable through any path according to the discovered topology

  • CLUSTERER_SEND_ERR - destination node or valid next hop appear to be reachable but send failed

2.1.7.  send_all(packet, cluster_id)

Send the given BIN packet to all the nodes in the specified cluster. The function operates similarly to send_to.

Meaning of the parameters is as follows:

  • bin_packet_t packet - the packet to be sent

  • int cluster_id - the cluster id

The function returns one of the following:

  • CLUSTERER_SEND_SUCCESS - successfully sent packet to at least one node

  • CLUSTERER_CURR_DISABLED - current node is disabled so sending is impossbile

  • CLUSTERER_DEST_DOWN - all nodes in the cluster are unreachable according to the discovered topology

  • CLUSTERER_SEND_ERR - send failed for all nodes in the cluster

2.1.8.  get_next_hop(cluster_id, node_id)

This function returns the next hop from the computed shortest path to the given destination node in the specified cluster. This is the node that is the actual destination for the send_to and send_all functions when the direct link with the intended destination is down. The function returns the same structure as get_nodes.

Meaning of the parameters is as follows:

  • int cluster_id - the cluster id

  • int node_id - the node id of the destination for which the next hop is returned.

2.1.9.  free_next_hop(next_hop)

This function will free the next hop returned by get_next_hop.

Meaning of the parameters is as follows:

  • clusterer_node_t *next_hop - next hop to be freed

2.1.10.  register_module(mod_name, cb, auth_check, accept_clusters_ids, no_accept_clusters)

This function registers an OpenSIPS module in order to receive BIN packets and cluster notifications. A certain module can accept packets from multiple clusters and provides a single callback function that will be called for each received packet. This function will also be called to notify cluster events like nodes becoming reachable/unreachable.

Meaning of the parameters is as follows:

  • char *mod_name - module name

  • clusterer_cb_f cb - callback function

  • int auth_check - 0 - no check, 1 - for every BIN packet received check if source IP belongs to one of the nodes in the cluster

  • int* accept_clusters_ids - array of cluster ids from which packets are accepted

  • int no_accept_clusters - length of accept_clusters_ids array

The callback function prototype:

...
typedef void (*clusterer_cb_f)(enum clusterer_event ev,bin_packet_t *, int packet_type,
                struct receive_info *ri, int cluster_id, int src_id, int dest_id);
...

Possble values for the event signaled through ev parameter of the callback funtion:

  • CLUSTER_RECV_MSG - received BIN message

  • CLUSTER_ROUTE_FAILED - failed to route a received BIN packet destined for another node in the cluster

  • CLUSTER_NODE_UP - a node became reachable

  • CLUSTER_NODE_DOWN - a node became unreachable

Chapter 3. Contributors

3.1. By Commit Statistics

Table 3.1. Top contributors by DevScore(1), authored commits(2) and lines added/removed(3)

 NameDevScoreCommitsLines ++Lines --
1. Vlad Patrascu (@rvlad-patrascu)2639199115202
2. Liviu Chircu (@liviuchircu)4835620439
3. Eseanu Marius Cristian (@eseanucristian)45103142534
4. Razvan Crainea (@razvancrainea)2317319143
5. Bogdan-Andrei Iancu (@bogdan-iancu)21101116108
6. Ionel Cerghit (@ionel-cerghit)82250212
7. Fabian Gast (@fgast)4233
8. Gohar Ahmed (@goharahmed)3165
9. Peter Lemenkov (@lemenkov)3111

(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. Regarding imported patches/code, we do our best to count the work on behalf of the proper owner, as per the "fix_authors" and "mod_renames" arrays in opensips/doc/build-contrib.sh. If you identify any patches/commits which do not get properly attributed to you, please submit a pull request which extends "fix_authors" and/or "mod_renames".

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

3.2. By Commit Activity

Table 3.2. Most recently active contributors(1) to this module

 NameCommit Activity
1. Vlad Patrascu (@rvlad-patrascu)Jul 2016 - Jan 2021
2. Fabian Gast (@fgast)Nov 2018 - Jul 2020
3. Liviu Chircu (@liviuchircu)Mar 2016 - Jul 2020
4. Bogdan-Andrei Iancu (@bogdan-iancu)Apr 2016 - Jun 2020
5. Razvan Crainea (@razvancrainea)Nov 2015 - Sep 2019
6. Gohar Ahmed (@goharahmed)Mar 2019 - Mar 2019
7. Peter Lemenkov (@lemenkov)Jun 2018 - Jun 2018
8. Ionel Cerghit (@ionel-cerghit)Dec 2016 - Dec 2016
9. Eseanu Marius Cristian (@eseanucristian)Jul 2015 - Sep 2015

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

Chapter 4. Documentation

4.1. Contributors

Last edited by: Fabian Gast (@fgast), Liviu Chircu (@liviuchircu), Vlad Patrascu (@rvlad-patrascu), Razvan Crainea (@razvancrainea), Bogdan-Andrei Iancu (@bogdan-iancu), Peter Lemenkov (@lemenkov), Eseanu Marius Cristian (@eseanucristian).

Documentation Copyrights:

Copyright © 2015-2017 www.opensips-solutions.com