Skip to main content
Version: 23.10

Installing a Centreon HA 2-nodes cluster

Prerequisites​

Understanding​

Before applying this procedure, you should have a good knowledge of Linux OS, of Centreon, and of the Pacemaker-Corosync clustering tools in order to have a proper understanding of what is being done and to be able to correct any mistakes that might occur.

WARNING: Anyone following this procedure does so at their own risk. Under no circumstances shall Centreon be liable for any breakdown or data loss.

Network Flows​

In addition to the necessary flows described in the official documentation, you will need to open the following flows:

FromDestinationProtocolPortApplication
Active NodePassive NodeSSHTCP 22Synchronization of configuration files (Must also be open from passive to active node)
Active NodePassive NodeBDDOTCP 5670RRDs synchronization (Must also be from passive to active node)
Active NodePassive NodeMySQLTCP 3306MySQL synchronization (Must also be open from passive to active node)
Central Servers + QDeviceCentral Servers + QDeviceCorosyncUDP 5404Communication inside the cluster (Multicast)
Central Servers + QDeviceCentral Servers + QDeviceCorosyncUDP 5405Communication inside the cluster (Unicast)
Central Servers + QDeviceCentral Servers + QDevicePCSTCP 2224Communication inside the cluster
Central Servers + QDeviceCentral Servers + QDeviceCorosyncTCP 5403Communication with the QDevice

Installed Centreon platform​

A Centreon HA cluster can only be installed on the base of an operating Centreon platform. Before following this procedure, it is mandatory that this installation procedure has already been completed and that about 5 GB of free space has been spared on the LVM volume group that carries the MariaDB data directory (/var/lib/mysql mount point by default).

The output of the vgs command must look like this: (pay attention to the value under VFree):

  VG                    #PV #LV #SN Attr   VSize   VFree 
  centos_centreon-c1      1   5   0 wz--n- <31,00g <5,00g
  • MariaDB files ibdata* and ib_logfile* must be in the "datadir" directory or in a subdirectory (scripts centreondb-smooth-backup.sh and mysql-sync-bigdb.sh are not compatible with this operation);
  • MariaDB files log-bin* and relay-log* can be located in a directory (or a subdirectory) different from "datadir". They can also be on a different logical volume (lvm) than "datadir". However, the logical volume must be located in the volume group where "datadir" is stored.

WARNING: If these particular prerequisites are not effective, the database synchronization method described below will not work.

Quorum Device​

In order to keep the cluster safe from split-brain issues, a third server is mandatory to resolve the master's election in the event of a connection loss. The role of Quorum Device can be held by a poller of the monitoring platform.

WARNING: Make sure Selinux and Firewalld are disabled.

Defining hosts' names and addresses​

In this procedure, we will refer to characteristics that are bound to change from one platform to another (such as IP addresses) by the following macros:

  • @CENTRAL_MASTER_IPADDR@: primary central server's IP address
  • @CENTRAL_MASTER_NAME@: primary central server's name (must be identical to hostname -s)
  • @CENTRAL_SLAVE_IPADDR@: secondary central server's IP address
  • @CENTRAL_SLAVE_NAME@: secondary central server's name (must be identical to hostname -s)
  • @QDEVICE_IPADDR@: quorum device's IP address
  • @QDEVICE_NAME@: quorum device's name (must be identical to hostname -s)
  • @MARIADB_REPL_USER@: MariaDB replication login (default: centreon-repl)
  • @MARIADB_REPL_PASSWD@: MariaDB replication password
  • @MARIADB_CENTREON_USER@: MariaDB Centreon login (default: centreon)
  • @MARIADB_CENTREON_PASSWD@: MariaDB Centreon password
  • @VIP_IPADDR@: virtual IP address of the cluster
  • @VIP_IFNAME@: network device carrying the cluster's VIP
  • @VIP_CIDR_NETMASK@: subnet mask length in bits (e.g. 24)
  • @VIP_BROADCAST_IPADDR@: cluster's VIP broadcast address
  • @CENTREON_CLUSTER_PASSWD@: hacluster user's password

Configuring centreon-broker​

On a standard Centreon platform, cbd service manages two processes of centreon-broker-daemon (cbd):

  • central-broker-master: also called "central broker" or "SQL broker", redirects input-output from pollers to database, RRD broker, and so on.
  • central-rrd-master: also called "RRD broker", receives the stream from the central broker and updates the RRD binary data files (used to display graphs).

In the context of a Centreon HA cluster, both broker processes will be handled by a separate service, managed by the cluster.

  • central-broker-master known as the resource cbd_central_broker, linked to systemd service cbd-sql
  • central-rrd-master known as the clone resource cbd_rrd, linked to systemd cbd service, the standard broker service of Centreon.

To ensure that everything goes well, you will have to unlink central-broker-master from the cbd service by checking "No" for parameter "Link to cbd service" in Configuration > Pollers > Broker configuration > central-broker-master under the General tab.

Double output stream toward RRD​

In the event of a cluster switch, you will expect the newly elected master central server to be able to display the metrics graphs, which requires all RRD data files to be up to date on both nodes. In order to fulfill this condition, you will double the central broker output stream and send it to both RRD broker processes. You can configure this in the same menu as above, this time under the Output tab. The parameters that must be changed are:

  • In the first "IPv4" output, replace "localhost" with @CENTRAL_MASTER_IPADDR@ in the "Host to connect to" field.
Output IPv4
Namecentreon-broker-master-rrd
Connection port5670
Host to connect to@CENTRAL_MASTER_IPADDR@
  • Add another "IPv4" output, similar to the first one, named "centreon-broker-slave-rrd" for example, directed toward @CENTRAL_SLAVE_IPADDR@.
Output IPv4
Namecentreon-broker-slave-rrd
Connection port5670
Host to connect to@CENTRAL_SLAVE_IPADDR@

Export the configuration​

Once the above actions have been done, export the central poller configuration files to apply these changes. Select the central poller and export the configuration with the "Move Export Files" option checked.

All the above actions should be applied either to both nodes or to @CENTRAL_MASTER_NAME@ only, and the exported files should be copied to @CENTRAL_SLAVE_NAME@:

rsync -a /etc/centreon-broker/*json @CENTRAL_SLAVE_IPADDR@:/etc/centreon-broker/

Customizing poller reload command​

You might not be aware that the central broker daemon must be reloaded every time you update your central poller's configuration, hence the "Centreon Broker reload command" parameter in Configuration > Pollers > Central.

As stated above, the centreon-broker processes will be divided into cbd (for RRD) and cbd-sql (for central broker) services. With this in mind, the service that needs to be reloaded is cbd-sql, and no longer cbd. So you will have to set the "Centreon Broker reload command" parameter to service cbd-sql reload.

System settings​

Before actually setting the cluster up, some system prerequisites must be satisfied.

Note : unless otherwise stated, each of the following steps must be performed on both central nodes.

Kernel network tuning​

In order to improve the reliability of the cluster, and since Centreon HA only supports IPv4, we recommend applying the following kernel settings to all your Centreon servers (including pollers):

cat >> /etc/sysctl.conf <<EOF
net.ipv6.conf.all.disable_ipv6 = 1
net.ipv6.conf.default.disable_ipv6 = 1
net.ipv4.tcp_retries2 = 3
net.ipv4.tcp_keepalive_time = 200
net.ipv4.tcp_keepalive_probes = 2
net.ipv4.tcp_keepalive_intvl = 2
EOF
systemctl restart NetworkManager

Name resolution​

So that the Centreon HA cluster can stay in operation in the event of a DNS service breakdown, all the cluster nodes must know each other by name without DNS, using /etc/hosts.

cat >/etc/hosts <<"EOF"
127.0.0.1 localhost localhost.localdomain localhost4 localhost4.localdomain4
@CENTRAL_MASTER_IPADDR@ @CENTRAL_MASTER_NAME@
@CENTRAL_SLAVE_IPADDR@ @CENTRAL_SLAVE_NAME@
@QDEVICE_IPADDR@ @QDEVICE_NAME@
EOF

From here, @CENTRAL_MASTER_NAME@ will be named the "primary server/node" and @CENTRAL_SLAVE_NAME@ the "secondary server/node". This designation is arbitrary; the two nodes will of course be interchangeable once the setup is complete.

Installing system packages​

Centreon offers a package named centreon-ha-web, which provides all the files and dependencies required by a Centreon cluster. These packages must be installed on both central nodes:

dnf config-manager --enable ha
dnf install centreon-ha-web pcs pacemaker corosync corosync-qdevice

SSH keys exchange​

SSH key-based authentication must be set so that files and commands can be sent from one node to another by UNIX accounts:

  • mysql
  • centreon

There are two ways of exchanging such keys:

  • By using the ssh-copy-id command: needs to be able to log in to remote host using a password. It is however unsafe for such system accounts to have a password authentication available. If you choose this method, we advise you to revoke the password afterward with these commands: passwd -d centreon and passwd -d mysql.
  • By manually copying the public key in ~/.ssh/authorized_keys. This method is safer.

The second method will be documented below.

centreon account​

Switch to centreon's bash environment on both nodes:

su - centreon

Then run these commands on both nodes:

ssh-keygen -t ed25519 -a 100
cat ~/.ssh/id_ed25519.pub

Once done, copy the content of the public key file displayed by cat and paste it to ~/.ssh/authorized_keys (must be created) on the other node, and then apply the correct file permissions (still as user centreon):

chmod 600 ~/.ssh/authorized_keys

The key exchange must be validated by an initial connection from each node to the other in order to accept and register the peer node's SSH fingerprint (still as user centreon):

ssh <peer node hostname>

Then exit the centreon session by typing exit or Ctrl-D.

mysql account​

For the mysql account, the procedure is slightly different because this user normally has neither a home directory nor the ability to open a Shell session. These commands must be run on both nodes as well:

systemctl stop mysql
mkdir /home/mysql
chown mysql: /home/mysql
usermod -d /home/mysql mysql
usermod -s /bin/bash mysql
systemctl start mysql
su - mysql

Once in mysql's bash envinronment, run these commands on both nodes:

ssh-keygen -t ed25519 -a 100
cat ~/.ssh/id_ed25519.pub

Once done, copy the content of the public key file displayed by cat and paste it to ~/.ssh/authorized_keys (must be created) on the other node, and then apply the correct file permissions (still as usermysql):

chmod 600 ~/.ssh/authorized_keys

The key exchange must be validated by an initial connection from each node to the other in order to accept and register the peer node's SSH fingerprint (still as user mysql):

ssh <peer node hostname>

Then exit the mysql session by typing exit or Ctrl-D.

Configuring the MariaDB database replication​

A Master-Slave MariaDB cluster will be set up so that everything is synchronized in real time.

Note: unless otherwise stated, each of the following steps must be performed on both central nodes.

Configuring MariaDB​

For both optimization and cluster reliability purposes, you need to add these tuning options to the MariaDB configuration in the /etc/my.cnf.d/server.cnf file. By default, the [server] section of this file is empty. Paste the following lines (some need to be modified) into this section:

[server]
server-id=1 # SET TO 1 FOR MASTER AND 2 FOR SLAVE
#read_only
log-bin=mysql-bin
binlog-do-db=centreon
binlog-do-db=centreon_storage
innodb_flush_log_at_trx_commit=1
sync_binlog=1
binlog_format=MIXED
slave_compressed_protocol=1
slave_parallel_mode=conservative
datadir=/var/lib/mysql
pid-file=/var/lib/mysql/mysql.pid
skip-slave-start
log-slave-updates
gtid_strict_mode=ON
expire_logs_days=7
ignore-db-dir=lost+found

# Tuning standard Centreon
innodb_file_per_table=1
open_files_limit=32000
key_buffer_size=256M
sort_buffer_size=32M
join_buffer_size=4M
thread_cache_size=64
read_buffer_size=512K
read_rnd_buffer_size=256K
max_allowed_packet=128M
# Uncomment for 4 Go Ram
#innodb_buffer_pool_size=512M
# Uncomment for 8 Go Ram
#innodb_buffer_pool_size=1G

Important: the value of server-id must be different from one server to the other. The values suggested in the comment 1 => Master and 2 => Slave are not mandatory, but recommended.

Reminder: Remember to uncomment the right value for innodb_buffer_pool_size according to your own servers' memory size.

To apply the new configuration, restart the database server:

systemctl restart mariadb

Make sure the restart went well:

systemctl status mariadb

Warning: Other files in /etc/my.cnf.d/, such as centreon.cnf, will be ignored from now on. Any customization will have to be added to server.cnf.

Warning: Remember to change the parameter Mysql configuration file path in Administration > Parameters > Backup

Securing the database server​

To avoid unnecessary exposure of your database, you should restrict access to it as much as possible. The mysql_secure_installation command will help you apply some basic security principles. You just need to run this command and let yourself be guided, choosing the recommended option at every step. We suggest you choose a strong password.

mysql_secure_installation

Creating the centreon MariaDB account​

First log in as root on both database servers (using the newly-defined password):

mysql -p

Then, on both sides, paste the following SQL commands to the MariaDB prompt to create the application user (default: centreon). Of course, you will replace the macros first:

CREATE USER '@MARIADB_CENTREON_USER@'@'@CENTRAL_SLAVE_IPADDR@' IDENTIFIED BY '@MARIADB_CENTREON_PASSWD@';
GRANT ALL PRIVILEGES ON centreon.* TO '@MARIADB_CENTREON_USER@'@'@CENTRAL_SLAVE_IPADDR@';
GRANT ALL PRIVILEGES ON centreon_storage.* TO '@MARIADB_CENTREON_USER@'@'@CENTRAL_SLAVE_IPADDR@';

CREATE USER '@MARIADB_CENTREON_USER@'@'@CENTRAL_MASTER_IPADDR@' IDENTIFIED BY '@MARIADB_CENTREON_PASSWD@';
GRANT ALL PRIVILEGES ON centreon.* TO '@MARIADB_CENTREON_USER@'@'@CENTRAL_MASTER_IPADDR@';
GRANT ALL PRIVILEGES ON centreon_storage.* TO '@MARIADB_CENTREON_USER@'@'@CENTRAL_MASTER_IPADDR@';

When upgrading to centreon-ha from an existing Centreon platform or an OVA/OVF VM deployment, update the '@MARIADB_CENTREON_USER@'@'localhost' password:

ALTER USER '@MARIADB_CENTREON_USER@'@'localhost' IDENTIFIED BY '@MARIADB_CENTREON_PASSWD@';

Creating the MariaDB replication account​

Still in the same prompt, create the replication user (default: centreon-repl):

GRANT SHUTDOWN, PROCESS, RELOAD, SUPER, SELECT, REPLICATION CLIENT, REPLICATION SLAVE ON *.* 
TO '@MARIADB_REPL_USER@'@'localhost' IDENTIFIED BY '@MARIADB_REPL_PASSWD@';

GRANT SHUTDOWN, PROCESS, RELOAD, SUPER, SELECT, REPLICATION CLIENT, REPLICATION SLAVE ON *.* 
TO '@MARIADB_REPL_USER@'@'@CENTRAL_SLAVE_IPADDR@' IDENTIFIED BY '@MARIADB_REPL_PASSWD@';

GRANT SHUTDOWN, PROCESS, RELOAD, SUPER, SELECT, REPLICATION CLIENT, REPLICATION SLAVE ON *.* 
TO '@MARIADB_REPL_USER@'@'@CENTRAL_MASTER_IPADDR@' IDENTIFIED BY '@MARIADB_REPL_PASSWD@';

Configuring the MariaDB scripts environment variables​

The /etc/centreon-ha/mysql-resources.sh file declares environment variables that must be configured so that the Centreon HA scripts dedicated to MariaDB can work properly. These variables must be assigned the chosen values for the macros.

#!/bin/bash

###############################
# Database access credentials #
###############################

DBHOSTNAMEMASTER='@CENTRAL_MASTER_NAME@'
DBHOSTNAMESLAVE='@CENTRAL_SLAVE_NAME@'
DBREPLUSER='@MARIADB_REPL_USER@'
DBREPLPASSWORD='@MARIADB_REPL_PASSWD@'
DBROOTUSER='@MARIADB_REPL_USER@'
DBROOTPASSWORD='@MARIADB_REPL_PASSWD@'
CENTREON_DB='centreon'
CENTREON_STORAGE_DB='centreon_storage'

###############################

To make sure that all the previous steps have been successful and that the correct names, logins and passwords have been entered in the configuration bash file, run this command:

/usr/share/centreon-ha/bin/mysql-check-status.sh

The expected output is:

Connection MASTER Status '@CENTRAL_MASTER_NAME@' [OK]
Connection SLAVE Status '@CENTRAL_SLAVE_NAME@' [OK]
Slave Thread Status [KO]
Error reports:
    No slave (maybe because we cannot check a server).
Position Status [SKIP]
Error reports:
    Skip because we can't identify a unique slave.

What matters here is that the first two connection tests are OK.

Switching to read-only mode​

Now that everything is well configured, enable the read_only on both database servers by uncommenting (i.e. removing the # at the beginning of the line) this instruction in the /etc/my.cnf.d/server.cnf file:

  • Primary node:
[server]
server-id=1
read_only
log-bin=mysql-bin
  • Secondary node:
[server]
server-id=2
read_only
log-bin=mysql-bin

Then apply this change by restarting MariaDB on both nodes:

systemctl restart mariadb

Synchronizing the databases and enabling MariaDB replication​

In the process of synchronizing the databases, you will first stop the secondary database process so that its data can be overwritten by the primary node's data.

Run this command on the secondary node:

systemctl stop mariadb

It is important to make sure that MariaDB is completely shut down. Run this command and check that it returns no output:

ps -ef | grep mariadb[d]

In case one or more processes are still alive, run this other command (it will prompt for the MariaDB root password):

mysqladmin -p shutdown

Once the service is stopped on the secondary node, run the synchronization script from the primary node:

/usr/share/centreon-ha/bin/mysql-sync-bigdb.sh

This script will perform the following actions:

  • checking that MariaDB is stopped on the secondary node
  • stopping MariaDB on the primary node
  • mounting an LVM snapshot on the same volume group that holds /var/lib/mysql (or whatever mount point holds the MariaDB data files)
  • starting MariaDB again on the primary node
  • recording the current position in the binary log
  • disabling the read_only mode on the primary node (this node will now be able to write to its database)
  • synchronizing/overwriting all the data files (except for the mysql system database)
  • unmounting the LVM snapshot
  • creating the replication thread that will keep both databases synchronized

This script's output is very verbose and you can't expect to understand everything, so to make sure it went well, focus on the last lines of its output, checking that it looks like this:

Umount and Delete LVM snapshot
  Logical volume "dbbackupdatadir" successfully removed
Start MySQL Slave
Start Replication
Id  User    Host    db  Command Time    State   Info    Progress
[variable number of lines]

The important thing to check is that Start MySQL Slave and Start Replication are present and are not followed by any errors.

In addition, the output of this command must display only OK results:

/usr/share/centreon-ha/bin/mysql-check-status.sh

The expected output is:

Connection MASTER Status '@CENTRAL_MASTER_NAME@' [OK]
Connection SLAVE Status '@CENTRAL_SLAVE_NAME@' [OK]
Slave Thread Status [OK]
Position Status [OK]

Setting up the Centreon cluster​

Note: unless otherwise stated, each of the following steps must be run on both central nodes.

Configuring the file synchronization service​

The file synchronization centreon-central-sync service needs the IP address of the peer node to be entered in its configuration file (/etc/centreon-ha/centreon_central_sync.pm).

So, on the @CENTRAL_MASTER_NAME@ server, the configuration file should look like this:

our %centreon_central_sync_config = (
    peer_addr => "@CENTRAL_SLAVE_IPADDR@"
);
1;

And on @CENTRAL_SLAVE_NAME@:

our %centreon_central_sync_config = (
    peer_addr => "@CENTRAL_MASTER_IPADDR@"
);
1;

Removing legacy Centreon cron jobs​

In a high-availability setup, gorgone daemon manages all cron-based scheduled tasks. To avoid cron on both nodes, remove all Centreon-related cron from the /etc/cron.d/ directory:

rm -f /etc/cron.d/centreon
rm -f /etc/cron.d/centstorage
rm -f /etc/cron.d/centreon-auto-disco
rm -f /etc/cron.d/centreon-ha-mysql

Permission modifications​

Modifications must be made to the permissions for the /var/log/centreon-engine and /tmp/centreon-autodisco directories.

In a clustered setup, it is a requirement to obtain a fully functional file sync and discovery scheduled task.

  • Files synchronization
chmod 775 /var/log/centreon-engine/
mkdir /var/log/centreon-engine/archives
chown centreon-engine: /var/log/centreon-engine/archives
chmod 775 /var/log/centreon-engine/archives/
find /var/log/centreon-engine/ -type f -exec chmod 664 {} \;
find /usr/share/centreon/www/img/media -type d -exec chmod 775 {} \;
find /usr/share/centreon/www/img/media -type f \( ! -iname ".keep" ! -iname ".htaccess" \) -exec chmod 664 {} \;
  • Services discovery
mkdir /tmp/centreon-autodisco/
chown apache: /tmp/centreon-autodisco/
chmod 775 /tmp/centreon-autodisco/

Stopping and disabling the services​

Centreon's application services will no longer be launched at boot time; they will be managed by the clustering tools. These services must therefore be stopped and disabled:

systemctl stop centengine snmptrapd centreontrapd gorgoned cbd httpd php-fpm centreon mysql
systemctl disable centengine snmptrapd centreontrapd gorgoned cbd httpd php-fpm centreon mysql

By default, the mysql service is enabled in both systemd and system V perspectives, so you should make sure it is disabled:

chkconfig mysql off

Creating the cluster​

Activating the clustering services​

First we enable all the services and start pcsd on both central nodes:

systemctl start pcsd

Preparing the server that will function as the quorum device​

You can use one of your pollers to play this role. It must be prepared with the commands below: 

dnf config-manager --enable ha
dnf install pcs corosync-qnetd
systemctl start pcsd.service
systemctl enable pcsd.service
pcs qdevice setup model net --enable --start
pcs qdevice status net --full

Modify the parameter COROSYNC_QNETD_OPTIONS in the file /etc/sysconfig/corosync-qnetd to make sure the service will be listening to the connections on Ipv4 only.

COROSYNC_QNETD_OPTIONS="-4"

Authenticating to the cluster's members​

For the sake of simplicity, the hacluster user will be assigned the same password on both central nodes and @QDEVICE_NAME@.

passwd hacluster

Now that both central nodes and the quorum device server share the same password, you will run this command only on one of the central nodes in order to authenticate on all the hosts taking part in the cluster.

pcs host auth \
    "@CENTRAL_MASTER_NAME@" \
    "@CENTRAL_SLAVE_NAME@" \
    "@QDEVICE_NAME@" \
    -u "hacluster" \
    -p '@CENTREON_CLUSTER_PASSWD@'

Creating the cluster​

The following command creates the cluster. It must be run only on one of the central nodes.

pcs cluster setup \
    centreon_cluster \
    "@CENTRAL_MASTER_NAME@" \
    "@CENTRAL_SLAVE_NAME@" \
    --force

Then start the pacemaker service on both central nodes:

systemctl enable pacemaker pcsd corosync
systemctl start pacemaker

And afterward define these properties only on one node:

pcs property set symmetric-cluster="true"
pcs property set stonith-enabled="false"
pcs resource defaults resource-stickiness="100"

You can now monitor the state of the cluster with the crm_mon -f command, which will display new resources as they appear.

Creating the Quorum Device​

Run this command on one of the central nodes:

pcs quorum device add model net \
    host="@QDEVICE_NAME@" \
    algorithm="ffsplit"

Creating the MariaDB cluster resources​

To be run only on one central node:

WARNING: the syntax of the following command depends on the Linux Distribution you are using.

pcs resource create "ms_mysql" \
    ocf:heartbeat:mariadb-centreon \
    config="/etc/my.cnf.d/server.cnf" \
    pid="/var/lib/mysql/mysql.pid" \
    datadir="/var/lib/mysql" \
    socket="/var/lib/mysql/mysql.sock" \
    binary="/usr/bin/mysqld_safe" \
    node_list="@CENTRAL_MASTER_NAME@ @CENTRAL_SLAVE_NAME@" \
    replication_user="@MARIADB_REPL_USER@" \
    replication_passwd='@MARIADB_REPL_PASSWD@' \
    test_user="@MARIADB_REPL_USER@" \
    test_passwd="@MARIADB_REPL_PASSWD@" \
    test_table='centreon.host'

WARNING: the syntax of the following command depends on the Linux Distribution you are using.

pcs resource promotable ms_mysql \
    master-node-max="1" \
    clone_max="2" \
    globally-unique="false" \
    clone-node-max="1" \
    notify="true"

Creating the clone resources​

Some resources must be running on only one node at a time (centengine, gorgone, httpd, ...), but some others can be running on both (the RRD broker and PHP8). For the second kind, you must declare clone resources.

Warning: All the commands in this chapter should be run only once on the central node of your choice.

PHP8 resource​
pcs resource create "php" \
    systemd:php-fpm \
    meta target-role="started" \
    op start interval="0s" timeout="30s" \
    stop interval="0s" timeout="30s" \
    monitor interval="5s" timeout="30s" \
    clone
RRD broker resource​
pcs resource create "cbd_rrd" \
    systemd:cbd \
    meta target-role="started" \
    op start interval="0s" timeout="90s" \
    stop interval="0s" timeout="90s" \
    monitor interval="20s" timeout="30s" \
    clone

Creating the centreon resource group​

VIP address​
pcs resource create vip \
    ocf:heartbeat:IPaddr2 \
    ip="@VIP_IPADDR@" \
    nic="@VIP_IFNAME@" \
    cidr_netmask="@VIP_CIDR_NETMASK@" \
    broadcast="@VIP_BROADCAST_IPADDR@" \
    flush_routes="true" \
    meta target-role="started" \
    op start interval="0s" timeout="20s" \
    stop interval="0s" timeout="20s" \
    monitor interval="10s" timeout="20s" \
    --group centreon
Httpd service​
pcs resource create http \
    systemd:httpd \
    meta target-role="started" \
    op start interval="0s" timeout="40s" \
    stop interval="0s" timeout="40s" \
    monitor interval="5s" timeout="20s" \
    --group centreon \
    --force
SQL Broker​
pcs resource create cbd_central_broker \
    systemd:cbd-sql \
    meta target-role="started" \
    op start interval="0s" timeout="90s" \
    stop interval="0s" timeout="90s" \
    monitor interval="5s" timeout="30s" \
    --group centreon
Gorgone service​
pcs resource create gorgone \
    systemd:gorgoned \
    meta target-role="started" \
    op start interval="0s" timeout="90s" \
    stop interval="0s" timeout="90s" \
    monitor interval="5s" timeout="20s" \
    --group centreon
centreon-central-sync service​

This service only exists in the context of Centreon HA. It provides real-time synchronization for configuration files, images, etc.

pcs resource create centreon_central_sync \
    systemd:centreon-central-sync \
    meta target-role="started" \
    op start interval="0s" timeout="90s" \
    stop interval="0s" timeout="90s" \
    monitor interval="5s" timeout="20s" \
    --group centreon
Centengine service​
pcs resource create centengine \
    systemd:centengine \
    meta multiple-active="stop_start" target-role="started" \
    op start interval="0s" timeout="90s" stop interval="0s" timeout="90s" \
    monitor interval="5s" timeout="30s" \
    --group centreon
Centreontrapd service​
pcs resource create centreontrapd \
    systemd:centreontrapd \
    meta target-role="started" \
    op start interval="0s" timeout="30s" \
    stop interval="0s" timeout="30s" \
    monitor interval="5s" timeout="20s" \
    --group centreon
Snmptrapd service​
pcs resource create snmptrapd \
    systemd:snmptrapd \
    meta target-role="started" \
    op start interval="0s" timeout="30s" \
    stop interval="0s" timeout="30s" \
    monitor interval="5s" timeout="20s" \
    --group centreon

Colocation constraints​

In order to force the cluster running both the centreon resource group and the MariaDB Master onto the same node, you must declare these colocation constraints:

pcs constraint colocation add master "centreon" with "ms_mysql-clone"
pcs constraint colocation add master "ms_mysql-clone" with "centreon"

After this step, all resources should be running on the same node; the platform should be redundant and working properly.

Activating the resources​

pcs resource enable php-clone
pcs resource enable cbd_rrd-clone
pcs resource meta vip target-role="started"
pcs resource meta centreontrapd target-role="started"
pcs resource meta snmptrapd target-role="started"
pcs resource meta centengine target-role="started"
pcs resource meta cbd_central_broker target-role="started"
pcs resource meta gorgone target-role="started"
pcs resource meta centreon_central_sync target-role="started"
pcs resource meta http target-role="started"

Checking the state of the cluster​

Checking the states of the resources​

You can monitor the cluster's resources in real time using the crm_mon -f command:

Cluster Summary:
  * Stack: corosync
  * Current DC: @CENTRAL_MASTER_NAME@ (version 2.0.5-9.0.1.el8_4.1-ba59be7122) - partition with quorum
  * Last updated: Wed Sep 15 16:35:47 2021
  * Last change:  Wed Sep 15 10:41:50 2021 by root via crm_attribute on @CENTRAL_MASTER_NAME@
  * 2 nodes configured
  * 14 resource instances configured
Node List:
  * Online: [ @CENTRAL_MASTER_NAME@ @CENTRAL_SLAVE_NAME@ ]
Full List of Resources:
  * Clone Set: ms_mysql-clone [ms_mysql] (promotable):
    * Masters: [ @CENTRAL_MASTER_NAME@ ]
    * Slaves: [ @CENTRAL_SLAVE_NAME@ ]
  * Clone Set: php-clone [php]:
    * Started: [ @CENTRAL_MASTER_NAME@ @CENTRAL_SLAVE_NAME@ ]
  * Clone Set: cbd_rrd-clone [cbd_rrd]:
    * Started: [ @CENTRAL_MASTER_NAME@ @CENTRAL_SLAVE_NAME@ ]
  * Resource Group: centreon:
    * vip       (ocf::heartbeat:IPaddr2):        Started @CENTRAL_MASTER_NAME@
    * http      (systemd:httpd):         Started @CENTRAL_MASTER_NAME@
    * cbd_central_broker        (systemd:cbd-sql):       Started @CENTRAL_MASTER_NAME@
    * gorgone   (systemd:gorgoned):      Started @CENTRAL_MASTER_NAME@
    * centreon_central_sync     (systemd:centreon-central-sync):         Started @CENTRAL_MASTER_NAME@
    * centengine        (systemd:centengine):    Started @CENTRAL_MASTER_NAME@
    * centreontrapd     (systemd:centreontrapd):         Started @CENTRAL_MASTER_NAME@
    * snmptrapd (systemd:snmptrapd):     Started @CENTRAL_MASTER_NAME@

If centreon_central_sync won't start, check if the folder /usr/share/centreon-broker/lua exists.

If not, you can create it with this command: mkdir -p /usr/share/centreon-broker/lua. And launch a cleanup with this command: pcs resource cleanup.

Disabled resources​

When you do a crm_mon -fr and you have a resource that is disabled:

...
 Master/Slave Set: ms_mysql-master [ms_mysql]
     Masters: [ @DATABASE_MASTER_NAME@ ]
     Slaves: [ @DATABASE_SLAVE_NAME@ ]
     Stopped: [ @CENTRAL_MASTER_NAME@ @CENTRAL_SLAVE_NAME@ ]
ms_mysql       (ocf::heartbeat:IPaddr2):       Stopped (disabled)
...

You must enable the resource with the following command:

pcs resource enable @RESSOURCE_NAME@

In our case:

pcs resource enable ms_mysql

Checking the database replication thread​

The MariaDB replication state can be monitored at any time with the mysql-check-status.sh command:

/usr/share/centreon-ha/bin/mysql-check-status.sh

The expected output is:

Connection MASTER Status '@CENTRAL_MASTER_NAME@' [OK]
Connection SLAVE Status '@CENTRAL_SLAVE_NAME@' [OK]
Slave Thread Status [OK]
Position Status [OK]

It can happen that the replication thread does not run properly after installation. Restarting the ms_mysql resource may fix it.

pcs resource restart ms_mysql-clone

Checking the constraints​

Normally, the two colocation constraints created during the setup should be the only constraints the pcs constraint command displays:

Location Constraints:
Ordering Constraints:
Colocation Constraints:
  centreon with ms_mysql-clone (score:INFINITY) (rsc-role:Started) (with-rsc-role:Master)
  ms_mysql-clone with centreon (score:INFINITY) (rsc-role:Master) (with-rsc-role:Started)
Ticket Constraints:

Integrating pollers​

You can now add your pollers and start monitoring!

?