Edit

Advanced configuration

Centreon Broker and the firewall

In certain cases you may not be able to initialize the Centreon Broker data flow from the poller (or the Remote Server) to the Central Server or the Remote Server.

Centreon has, however, developed a solution for initializing the flow from the Centreon Central Server, or from the Remote Server, to the poller.

Go to the Configuration > Pollers > Broker configuration menu and click on Centreon Broker SQL configuration on the Central Server or Remote Server.

Go to the Input tab panel and add a new TCP - IPv4 entry.

Enter the Name of the configuration, the TCP Connection port for connecting to the poller, and the Host to connect to. Then Save your configuration.

Go to the Configuration > Pollers > Broker configuration menu and click on the Broker module of your poller.

On the Output tab panel modify the Output 1 - IPv4 form:

1. Remove the entry for Host to connect to.
2. Check the Connection port.
3. Set Yes for One peer retention option.

Click Save and generate the configuration of the affected servers.

Centreon Broker flow authentication

If you need to authenticate pollers that are sending data to the monitoring system, you can use the Centreon Broker authentication mechanism, which is based on X.509 certificates.

First generate a Certificate Authority (CA) certificate with OpenSSL. ca.key will be the private key (stored securely), while ca.crt will be the public certificate for authenticating incoming connections:

openssl req -x509 -newkey rsa:2048 -nodes -keyout ca.key -out ca.crt -days 365

Now generate the certificates using the CA key:

openssl req -new -newkey rsa:2048 -nodes -keyout central.key -out central.csr -days 365
openssl req -new -newkey rsa:2048 -nodes -keyout poller.key -out poller.csr -days 365
openssl x509 -req -in central.csr -CA ca.crt -CAkey ca.key -CAcreateserial -out central.crt -days 365 -sha256
openssl x509 -req -in poller.csr -CA ca.crt -CAkey ca.key -CAcreateserial -out poller.crt -days 365 -sha256

Put central.key, central.crt and ca.crt on the Centreon central server (e.g., in /etc/centreon-broker) and poller.key, poller.crt and ca.crt on your poller.

You must now configure the Centreon Broker in order to use these files. Go to Configuration > Pollers > Broker configuration. For the central-broker-master, under the Input tab, set the following parameters for central-broker-master-input:

• Enable TLS encryption = Yes
• Private key file = /etc/centreon-broker/central.key
• Public certificate = /etc/centreon-broker/central.crt
• Trusted CA's certificate = /etc/centreon-broker/ca.crt

As with the poller, you will have to modify the TCP output under the Output tab with the following parameters:

• Enable TLS encryption = Yes
• Private key file = /etc/centreon-broker/poller.key
• Public certificate = /etc/centreon-broker/poller.crt
• Trusted CA's certificate = /etc/centreon-broker/ca.crt

Regenerate the configuration of the affected pollers (Configuration > Pollers > Pollers) and the authentication will be complete.

Centreontrapd Configuration

Poller

You must edit the Centreontrapd configuration file to be able to use the local SQLite database. Refer to the chapter Passive Monitoring.

Remote Server

Configuring the Centreontrapd process is the same as on the Central Server.

Advanced configuration of Centreon Broker

This section will help you understand how Centreon Broker works and how it should be configured according to Centreon's best practices. The various options used by Centreon Broker are described.

Overview

The Centreon Broker's core is a simple multiplexing engine that takes inputs events and sends them to various outputs. Inputs are typically other Centreon Broker instances received via TCP/IP, while outputs can be an SQL database, other brokers, a BI/BAM engine, Centreon Map, etc.

Each input or output has a type that describes what it does plus several parameters, some mandatory and others optional. Additionally, an output can have a failover that will start when the output is in an error state in order to ensure data retention.

An important distinction should be made between a standalone Centreon Broker and a Centreon Broker installed as a Centreon Engine module. Both have the same capabilities and support the same inputs and outputs. The difference is that the Centreon Broker configured as a module is automatically started when Centreon Engine starts. This broker automatically generates the events associated with the Centreon Engine. Broker modules often only have one output to a Centreon Broker instance acting as a concentrator.

Main configuration page

This section lists all the Centreon Broker instances configured in your infrastructure, either in standalone or module mode. Each instance has a name, is associated with a poller, has a number of inputs, outputs*, and loggers, and can be enabled or disabled.

A central-type poller will have three Centreon Broker instances by default:

• one Centreon Broker installed as a module for a Centreon Engine (called a central-module-master)
• one Centreon Broker acting as a stand-alone concentrator (called a central-broker-master)
• one Centreon Broker specialized in generating the RRD data used by graphs (called central-rrd-master).

The best practice is to always use a separate Centreon Broker instance to generate RRD data. This way, an issue occuring in the RRD stack will not have any impact on your main monitoring.

As expected, the central-module-master has only one output and zero inputs. Configured as a Centreon Engine module, it generates events on its own and forwards them to the standalone Centreon Broker instance.

A poller generally has only one Centreon Broker instance configured as a Centreon Engine module.

Broker general configuration page

This section lists all the general options associated with a Centreon Broker instance.

Main options:

• Poller
  The poller containing the instance.

• Name
  The name of the instance.

• Config file name
  The name of the configuration file used by this instance.

• Retention path
  When an output is in an error state, a failover is launched. Failovers save data in files called retention files. These in turn are saved in the directory specified here. The best practice is /var/lib/centreon-broker/.

• Status
  Used to enable or disable the instance.

Log options:

• Write timestamp
  If activated, each log entry is preceded by the timestamp of the time it was written. This is useful to know when an error has occured. Best practice is Yes.

• Write thread id
  If activated, each log entry is preceded by the ID of the thread being executed at that instant. This is only used for advanced debugging purposes. Best practice is No.

Advanced Options:

• Statistics
  Centreon Broker has an on-demand status reporting mechanism that can be enabled here. This is used by Centreon Web to check the status of the instance at any time and determine which inputs and outputs are in an error state and to generate various statistics on event processing. Best practice is Yes.

• Correlation
  Centreon Broker has a top-level correlation mechanism. This should only be activated if top-level correlation has been properly configured in Centreon Web. In all other cases, default is No.

• Event queue max size
  The maximum size of the in-memory queue in events. If the number of events in memory exceeds this number, Centreon Broker will start to use temporary files to prevent the broker from using too much memory. This, however, causes additional disk I/O. The exact number can be adjusted to use more or less memory. A suggested default is 50000.

If Statistics is enabled, on-demand status can be queried manually through a file in /var/lib/centreon-broker/name.stats.

Broker input configuration page

This section lists all the inputs activated for this instance of Centreon Broker. Centreon Broker can have as many inputs as needed.

Inputs read events from a TCP connection. All inputs have the following parameters:

• Name
  The name of the input. Must be unique.

• Serialization protocol
  The protocol that was used to serialize the data. Can be either BBDO or NDO. NDO is a legacy textual protocol with inferior performance, data density and security. BBDO is a next-generation binary protocol that is effective and secure. NDO is deprecated. It should never be used for a new software installation. Best practice is BBDO.

• Compression
  If compression is used to serialize the data, the options are: auto, yes, or no. If left on auto, the Centreon Broker will detect whether compression was used during a TCP handshake (or assume that no compression was used for files). Default is auto for TCP, no for files.

• Filter category
  The categories of events accepted by this input. If empty, no restriction on events accepted. If filled, only events of the given type will be processed. Inputs that accept data from the Centreon Engine Broker module should be set to only accept Neb events.

• Connection Port
  The port that will be used for the connection. Mandatory.

• Host to connect to
  This important parameter decides whether the input will listen or attempt to initiate a connection. If left empty, the input will listen on its given port. If specified, it will attempt to initiate a connection to the given host/port.

• Enable TLS encryption
  Enables the encryption of the flow. For the encryption to work, the private key file, the Public certificate and the Trusted CA's certificate need to be set on both ends. Default is auto, i.e., no unless TCP negotiation has been activated and the remote endpoint has activated encryption.

• Private Key File
  The private key file used for the encryption.

• Public certificate
  The public certificate used for the encryption.

• Trusted CA's certificate
  The trusted CA certificate used for the encryption.

• Enable negotiation
  If set to yes, this input will try to negotiate encryption and compression with the remote endpoint.

• One peer retention mode
  By default, a listening input will accept any number of incoming connections. In one peer retention mode only one connection at a time is accepted, on a first-come first-serve basis. Default is no.

TCP input can either listen on a given port or can attempt to initiate a connection if a host is given. This allows flexible network topology.

Broker Logger configuration page

This section lists all the loggers activated for this Centreon Broker instance. A Centreon Broker can have as many loggers as needed.

For each logger, the parameters are the following:

• Type
  Four types of loggers are managed by Centreon Broker:

  1. File: This logger will write a log into the file specified in its name parameter.
  2. Standard: This logger will write into the standard output if named stdout or cout or into the standard error output if named stderr or cerr.
  3. Syslog: This logger will write into the syslog as provided by the system, prefixed by centreonbroker.
  4. Monitoring: This logger will write in the Centreon Engine log. It should only be activated if the Centreon Broker instance is loaded by the Centreon Engine at start-up.

• Name
  The name of this logger. This name must be the path of a file if the logger has the type file and either stdout, cout, stderr or cerr, if the logger has the type standard. This option is mandatory.

• Configuration messages
  Should configuration messages be logged? Configuration messages are one-time messages that pop up when Centreon Broker is started. Default is Yes.

• Debug messages
  Should debug messages be logged? Debug messages are messages used to debug Broker behavior. They are extremely verbose and should not be used in a production environment. Default is No.

• Error messages
  Should error messages be logged? Error messages are messages logged when a runtime error occurs. They are generally important. Default is Yes.

• Informational messages
  Should informational messages be logged? Informational messages are used to provide information on a specific subject. They are somewhat verbose. Default is No.

• Logging level
  The level of the verbosity accepted by this logger. The higher the verbosity, the more messages will be logged. Default is Base.

Additionally, the File type has the following parameter:

• Max file size
  The maximum size of a log file in bytes. When the file has reached its limit, the old data will be overwritten in a round-robin fashion.

A broker will usually have at least one file logger which will log configuration and error messages. Others can be configured freely. A maximal logger (every category set to Yes and logging level set to Very detailed) is valuable to debug some issues, but be warned that it will quickly generate a very large amount of data.

Broker output configuration page

This section lists all the outputs activated for this Centreon Broker instance. Centreon Broker can have as many outputs as needed.

For each output, the parameters are:

• Type
  There are several types of outputs managed by the Centreon Broker:

  1. TCP - IPV4 and TCP - IPV6: This output forwards data to another server, another Centreon Broker or Centreon Map.
  2. File: Writes data into a file.
  3. RRD: Generates RRD data from performance data.
  4. Storage: Writes metrics into the database and generates performance data.
  5. SQL: Writes the real-time status into Centreon's database.
  6. Dumper Reader: Reads from a database when Broker is asked to synchronize databases.
  7. Dumper Writer: Writes into a database when Broker is asked to synchronize databases.
  8. BAM Monitoring: Generates BAM data from raw events and updates real-time BAM status.
  9. BAM Reporting: Writes long-term BAM logs that can then be used by BI.

• Failover
  A failover is an output that will be started when in an error state. Examples are TCP connections "gone haywire" or a MySQL server suddenly disconnecting, etc. By default, each output has an automatic failover that will always store data in retention files and replay it when the primary output recovers from its error state. This is desirable 99% of the time. Alternatively, you can specify another output that will act as a failover if needed.

• Retry interval
  When the output is in an error state, this parameter controls the amount of time the output will wait before retrying. Default is one attempt every 30 seconds.

• Buffering timeout
  When this output is in an error state, Centreon Broker will wait a specified time before launching the failover. This is mainly useful if Centreon Broker should wait for another software to initialize before activating its failover. In all other cases, this parameter should not be used. Default is 0 seconds.

• Filter category
  The categories of events accepted by this output. If left empty, no restriction on events accepted. If filled, only events of the given type will be processed. The exact best practices are output specific:

  1. BAM Reporting should only accept BAM events.
  2. Dump Writer should only accept dumper events.
  3. RRD should only accept storage events.

  In all other cases, no restriction should be configured.

Events generated by an output are reinjected into Centreon Broker's event queue.

Some outputs only work when consuming data generated by another output. An RRD output consumes data from a storage output, a dumper writer output consumes data from a dumper reader, and a BAM reporting output consumes data from a BAM monitoring output.

Centreon Web needs at least an active output SQL ouput to activate its real-time monitoring capabilities. The storage and RRD outputs are needed to activate Centreon Web metric plotting. The BAM monitoring output is needed for real-time BAM data and the BAM reporting output for BI reports.

Due to the fully distributed nature of Centreon Broker, producer and consumer outputs can be located on logically or physically different instances as long as they are connected to each other.

Important: Centreon Web 2.x features two databases, the configuration database and the real-time database. Those are respectively called centreon and centreon-storage. Different outputs expect may different databases in their configuration.

Output Type	Expected database
SQL	centreon-storage
Storage	centreon-storage
Dumper Reader	centreon
Dumper Writer	centreon
BAM Monitoring	centreon
BAM Reporting	centreon-storage

TCP outputs

TCP outputs forward events to a remote endpoint. As with TCP inputs, TCP outputs can either listen on a given port or attempt to initiate a connection if a host parameter is given. This allows for flexible network topology.

TCP-type outputs have the following parameters:

• Serialization protocol
  The protocol used to serialize the data. Can be either BBDO or NDO. NDO is an legacy textual protocol with inferior performance, data density and security. BBDO is a next-generation binary protocol that is effective and secure. NDO is deprecated. It should never be used for new installations. Best practice is BBDO.

• Enable negotiation
  If yes, this output will try to negotiate encryption and compression with the remote endpoint.

• Connection Port
  Port used for the connection. Mandatory.

• Host to connect to
  This key parameter decides whether the input will listen or attempt to initiate a connection. If left empty, the input will listen on its given port. If specified, it will attempt to initiate a connection to the given host/port.

• Enable TLS encryption
  Enables the encryption of the flow. For the encryption to work, the private key file, Public certificate and Trusted CA's certificate need to be set on both ends. Default is auto, i.e., no unless TCP negotiation has been activated and the remote endpoint has activated encryption.

• Private Key File
  The private key file used for the encryption.

• Public certificate
  The public certificate used for the encryption.

• Trusted CA's certificate
  The trusted CA certificate used for the encryption.

• One peer retention mode
  By default, a listening input will accept any number of incoming connections. In one peer retention mode only one connection at a time is accepted, on a first-come first-serve basis. Default is no.

• Compression
  If compression should be used to serialize the data. Can be auto, yes, or no. If left on auto Centreon Broker will detect if compression is supported by the endpoint during a TCP negotiation. Default is auto for TCP.

• Compression Level
  The level of compression that should be used, from 1 to 9. Default (or if not filled) is 6. The higher the compression level is, the higher the compression will be at the expense of processing power.

• Compression Buffer
  The size of the compression buffer that should be used. Best practice is 0 or nothing.

File outputs

File outputs send events into a file on the disk. Additionally, they have the capability of replaying the data of this file if used as a failover output. Most file outputs will be used as failovers.

File type outputs have the following parameters:

• Serialization protocol
  The protocol that was used to serialize the data. Can be either BBDO or NDO. NDO is a legacy textual protocol with inferior performance, data density and security. BBDO is a next-generation binary protocol that is effective and secure. NDO is deprecated. It should never be used for new installations. Best practice is BBDO.

• File path
  The path of the file being written to.

• Compression
  If compression should be used to serialize the data. Can be auto, yes, or no. auto is equal to no for files.

• Compression Level
  The level of compression to be used, from 1 to 9. Default (or if not filled) is 6. The higher the compression level is, the higher the compression will be at the expense of processing power.

• Compression Buffer
  The size of the compression buffer to be used. Best practice is 0.

RRD outputs

RRD outputs generate RRD data (used by Centreon Web to generate graphs) from metrics data generated by a storage output. The best practice is to isolate this output on its own Centreon Broker instance to ensure that an issue in the RRD stack will not have any impact on the main Centreon Broker instance.

RRD-type outputs have the following parameters:

• RRD file directory for metrics
  The directory where the RRD files of the metrics will be written. A recommended default is /var/lib/centreon/metrics/.

• RRD file directory for status
  The directory where the RRD files of the status will be written. A recommended default is /var/lib/centreon/status/

• TCP port
  The port used by RRDCached, if RRDCached has been configured on this server. If not, leave empty.

• Unix socket
  The Unix socket used by RRDCached, if RRDCached has been configured on this server. If not, leave empty.

• Write metrics
  Should RRD metric files be written? Default is yes.

• Write status
  Should RRD status files be written? Default is yes.

Storage Outputs

Perfdata storage outputs save metric data into a database and generate RRD data used by the RRD output. This output usually generates multiple queries and is very performance intensive. If Centreon Broker is slow, try adjusting the maximum queries per transaction parameter to optimize processing speed.

This output can be tasked to rebuild RRD data from a database of stored metric data. This is usually a slow, costly process, though you can simultaneously process new metric data at a reduced speed.

Storage-type outputs have the following parameters:

• DB Type
  The type of the database being accessed. MariaDB is a state-of-the-art database that has been extensively tested by Centreon. We advise using MariaDB.

• DB Port
  The port of the database being accessed.

• DB User
  The user account for connecting to this database.

• DB Name
  The name of the database. In Centreon terms, this is the database containing the real-time monitoring data, generally called centreon-storage.

• DB Password
  The password used by the output to connect to this database.

• Maximum queries per transaction
  This parameter is used to batch several queries in large transactions. This allows for improved performance but can generate latency if an insufficient number of queries are generated to fill those batches. The default is 20000 queries per transaction. If you have a low load and unexpectedly high latency, try lowering this number. If you have a high load and high latency, try raising it.

• Transaction commit timeout
  Number of seconds allowed before a forced commit is made. Default is infinite. If you have a low load and unexpectedly high latency, try 5 seconds.

• Replication enabled
  Should Centreon Broker check that the replication status of this database is complete before trying to insert data? Only useful if replication is enabled for this database.

• Rebuild check interval in seconds
  The number of seconds between each rebuild check. Default 300 seconds.

• Store in performance data in data_bin
  Should this output save the metric data in the database? Default is yes. If no, this output will generate RRD data without saving them into the database, making a rebuild impossible.

• Insert in index data
  Should new index data be inserted into the database? Default is no. This should never be modified unless prompted by Centreon Support or explicitly advised in the documentation.

SQL outputs

SQL outputs save real-time status data into the real-time database used by Centreon Web. This is the most important output for the operation of Centreon Web.

Moreover, this output has a garbage collector that will clean old data from the database occasionally. This is an optional process, as old data is marked disabled, and can actually be useful to keep for debugging purpose.

SQL-type outputs have the following parameters:

• DB Type
  The type of the database being accessed. MariaDB is a state-of-the-art database that has been extensively tested by Centreon. We advise using MariaDB.

• DB Port
  The port of the database being accessed.

• DB User
  The user account for connecting to this database.

• DB Name
  The name of the database. In Centreon terms, this is the database containing the real-time monitoring data, generally called centreon-storage.

• DB Password
  The password used by the output to connect to this database.

• Maximum queries per transaction
  This parameter is used to batch several queries in large transactions. This allows for improved performance but can generate latency if an insufficient number of queries are generated to fill those batches. The default is 20000 queries per transaction. If you have a low load and unexpectedly high latency, try lowering this number. If you have a high load and high latency, try raising this number.

• Transaction commit timeout
  Number of seconds allowed before a forced commit is made. Default is infinite. If you have a low load and unexpectedly high latency, try 5 seconds.

• Replication enabled
  Should Centreon Broker check that the replication status of this database is complete before trying to insert data? Only useful if replication is enabled for this database.

• Cleanup check interval
  Number of seconds between each run of the garbage collector "cleaning" out old data in the database. Default is never.

• Instance timeout
  Number of seconds before an instance is marked as unresponding and all of its hosts and services marked as unknown. Default is 300 seconds.

Lua outputs

Lua outputs send metrics information into a script by a key-value system. The Lua script should reside on your server.

• Path
  The path of the Lua script in your server.

• Filter category
  The categories of events accepted by this output. If empty, no restriction on events is accepted. If specified, only events of the given type will be processed. Outputs that accept data from Centreon Engine's Broker module should be set to only accept Neb events.

Lua parameter

• Type
  Type of metric value.

• Name/Key
  Name of metric value.

• Value
  Value of metric.

Dumper reader/writer

A dumper reader/writer pair is used to synchronize part of a database between two instances of Centreon Broker. In the future we will provide an extensive synchronization mechanism, but today this system is mainly used to synchronize BAs for the BAM Poller Display mechanism.

The BAM Poller Display configuration documentation explains how to properly configure these outputs.

Dumper Reader-type and Dumper Writer-type outputs have the following parameters:

• DB Type
  The type of the database being accessed. MariaDB is a state-of-the-art database that has been extensively tested by Centreon. We advise using MariaDB.

• DB Port
  The port of the database being accessed.

• DB User
  The user account for connecting to this database.

• DB Name
  The name of the database. In Centreon terms, this is the database containing the real-time monitoring data, generally called centreon-storage.

• DB Password
  The password used by the output to connect to this database.