8. Installare e configurare Pgpool
Pianificazione dell'installazione:
Pgpool è responsabile dell'organizzazione di PostgreSQL descritta di seguito.
• Indirizzare il traffico in scrittura (ThingWorx) del client al nodo master PostgreSQL.
• Indirizzare il traffico in lettura del client al nodo di standby PostgreSQL.
• Gestione di tutti i nodi PostgreSQL durante diversi scenari di failover.
Pertanto, è fondamentale stabilire una relazione di trust tra i nodi Pgpool-II e i nodi PostgreSQL.
In questo esempio l'approccio per stabilire la relazione di trust è:
• Installare il componente PostgreSQL nel nodo Pgpool, verrà creato anche l'utente postgres.
• Impostare la stessa password per l'utente postgres in tutti i nodi: nodo master Pgpool, nodo secondario Pgpool, node0 PostgreSQL, node1 PostgreSQL e node2 PostgreSQL
• Creare la chiave SSH nel nodo master Pgpool e distribuire la chiave pubblica in tutti i nodi.
• Distribuire la chiave privata al nodo di standby Pgpool.
• Impostare una relazione di trust per l'utente postgres utente in tutti i nodi.
Passi principali per ottenere il trust precedente:
• ssh-keygen per generare idrsa e idrsa.pub nel nodo master Pgpool per l'utente postgres.
• Copiare il contenuto di idrsa.pub nella cartella ~/.ssh/authorizedkeys dell'utente postgres in tutti i nodi.
• chmod 700 ~/.ssh su tutti i nodi per l'utente postgres.
• vim /etc/ssh/sshd_config con il contenuto seguente:
PermitEmptyPasswords yes
PasswordAuthentication yes
PasswordAuthentication yes
• Impostare la password utente postgres sullo stesso valore per tutti i nodi.
• Testare ssh postgres@remote_IP per verificare che tutti i nodi Pgpool-II possano eseguire l'autenticazione in tutti i nodi PostgreSQL utilizzando la chiave SSH.
Installazione di Pgpool
In RHEL Pgpool può essere installato con i comandi seguenti:
sudo rpm -ivh http://www.pgpool.net/yum/rpms/3.6/redhat/rhel-6-x86_64/pgpool-II-release-3.6-1.noarch.rpm
sudo yum -y install pgpool-II-pg94
sudo yum -y install pgpool-II-pg94-debuginfo
sudo yum -y install pgpool-II-pg94-devel
sudo yum -y install pgpool-II-pg94-extensions
sudo yum -y install postgresql-contrib
sudo yum -y install pgpool-II-pg94
sudo yum -y install pgpool-II-pg94-debuginfo
sudo yum -y install pgpool-II-pg94-devel
sudo yum -y install pgpool-II-pg94-extensions
sudo yum -y install postgresql-contrib
Configurazione di pgpool
Eseguire questi passi in ogni nodo Pgpool-II.
Passo 1
Aggiornare pgpool.conf.
Accedere come utente pgpool.
Modificare pgpool.conf:
vim /etc/pgpool-II/pgpool.conf
con il contenuto seguente:
|
Nodo server
|
Configuration Item
|
Valore configurazione
|
Nodo
|
|---|---|---|---|
|
1
|
listen_address
|
*
|
A seconda del requisito di protezione, è possibile modificarlo in un indirizzo IP specifico.
|
|
2
|
port
|
5432
|
|
|
3
|
pcp_listen_addresses
|
*
|
A seconda del requisito di protezione, è possibile modificarlo in un indirizzo IP specifico.
|
|
4
|
pcp_port
|
9898
|
|
|
5
|
backend_hostname0
|
'10.91.9.200'
|
|
|
6
|
backend_port0
|
5432
|
|
|
7
|
backend_weight0
|
1
|
|
|
8
|
backend_data_directory0
|
'/db/postgres'
|
|
|
9
|
backend_flag0
|
ALLOWTOFAILOVER
|
|
|
10
|
backend_hostname1
|
'10.91.9.24'
|
|
|
11
|
backend_port1
|
5432
|
|
|
12
|
backend_weight1
|
1
|
|
|
13
|
backend_data_directory1
|
'/db/postgres'
|
|
|
14
|
backend_flag1
|
'ALLOWTOFAILOVER'
|
|
|
15
|
backend_hostname2
|
'10.91.9.41'
|
|
|
16
|
backend_port2
|
5432
|
|
|
17
|
backend_weight2
|
1
|
|
|
18
|
backend_data_directory2
|
'/db/postgres'
|
|
|
19
|
enable_pool_hba
|
on
|
|
|
20
|
master_slave_mode
|
on
|
|
|
21
|
master_slave_sub_mode
|
'stream'
|
|
|
22
|
sr_check_period
|
10
|
|
|
23
|
sr_check_user
|
'replicator'
|
|
|
24
|
sr_check_password
|
'replicator'
|
|
|
25
|
sr_check_database
|
'postgres'
|
|
|
26
|
health_check_user
|
'postgres'
|
|
|
27
|
health_check_password
|
'postgres'
|
|
|
28
|
health_check_database
|
'postgres'
|
|
|
29
|
failover_command
|
'/etc/pgpool-II/failover.sh %d %h %D %m %H %R %M %P'
|
 |
I parametri per la configurazione del watchdog non sono ancora inclusi in questo esempio.
|
Passo 2
Creare lo script di failover.
vim /etc/pgpool-II/failover.sh
con il contenuto seguente:
#!/bin/bash -x
echo ------------------------Begin failover------------------------
#------------------------------------
# Inputs
#------------------------------------
FAILED_NODEID=$1 # %d = node id
FAILED_HOSTNAME=$2 # %h = hostname
FAILED_DBPATH=$3 # %D = database cluster path
NEW_MASTER_NODEID=$4
NEW_MASTER_HOSTNAME=$5
NEW_MASTER_DBPATH=$6
OLD_MASTER_NODEID=$7
OLD_PRIMARY_NODEID=$8
# %m = new master node id
# %H = new master host name
# %R = new master database cluster path
# %M = old master node id
# %P = old primary node id
#------------------------
# Nodes
#------------------------
PGDATA=/db/postgres
PGBIN=/usr/pgsql-10.x/bin
LOCALBIN=/db/bin
NODE0_HOSTNAME=10.91.9.200
NODE0_ARCHIVE=/db/node0archive
NODE1_HOSTNAME=10.91.9.24
NODE1_ARCHIVE=/db/node1archive
NODE2_HOSTNAME=10.91.9.41
NODE2_ARCHIVE=/db/node2archive
if [ $FAILED_NODEID = $OLD_PRIMARY_NODEID ]; then
if ssh -T postgres@$NEW_MASTER_HOSTNAME grep -q "$FAILED_HOSTNAME" /db/postgres/recovery.conf; then
#yes, it is, go ahead to promote
echo Node $NEW_MASTER_HOSTNAME can be promoted safely
else
#otherwise, new master has to be switched.
if [ $FAILED_NODEID = 1 ]; then
if [ $NEW_MASTER_NODEID = 0 ]; then
NEW_MASTER_HOSTNAME=$NODE2_HOSTNAME
NEW_MASTER_NODEID=2
else
NEW_MASTER_HOSTNAME=$NODE0_HOSTNAME
NEW_MASTER_NODEID=0
fi;
fi;
if [ $FAILED_NODEID = 2 ]; then
if [ $NEW_MASTER_NODEID = 0 ]; then
NEW_MASTER_HOSTNAME=$NODE1_HOSTNAME
NEW_MASTER_NODEID=1
else
NEW_MASTER_HOSTNAME=$NODE0_HOSTNAME
NEW_MASTER_NODEID=0
fi;
fi;
if [ $FAILED_NODEID = 0 ]; then
if [ $NEW_MASTER_NODEID = 1 ]; then
NEW_MASTER_HOSTNAME=$NODE2_HOSTNAME
NEW_MASTER_NODEID=2
else
NEW_MASTER_HOSTNAME=$NODE1_HOSTNAME
NEW_MASTER_NODEID=1
fi;
fi;
echo New master has been switched to $NEW_MASTER_HOSTNAME
fi;
TARGET_HOSTNAME=0.0.0.0
TARGET_ARCHIVE=/db/archive
STANDBY1_HOSTNAME_TO_RETARGET=0.0.0.0
if [ $FAILED_NODEID = 0 ]; then
if [ $NEW_MASTER_NODEID = 1 ]; then
TARGET_HOSTNAME=$NEW_MASTER_HOSTNAME
TARGET_ARCHIVE=/db/node1archive
STANDBY1_HOSTNAME_TO_RETARGET=$NODE2_HOSTNAME
else
TARGET_HOSTNAME=$NEW_MASTER_HOSTNAME
TARGET_ARCHIVE=/db/node2archive
STANDBY1_HOSTNAME_TO_RETARGET=$NODE1_HOSTNAME
fi;
fi;
if [ $FAILED_NODEID = 1 ]; then
if [ $NEW_MASTER_NODEID = 0 ]; then
TARGET_HOSTNAME=$NEW_MASTER_HOSTNAME
TARGET_ARCHIVE=/db/node0archive
STANDBY1_HOSTNAME_TO_RETARGET=$NODE2_HOSTNAME
else
TARGET_HOSTNAME=$NEW_MASTER_HOSTNAME
TARGET_ARCHIVE=/db/node2archive
STANDBY1_HOSTNAME_TO_RETARGET=$NODE0_HOSTNAME
fi;
fi;
if [ $FAILED_NODEID = 2 ]; then
if [ $NEW_MASTER_NODEID = 0 ]; then
TARGET_HOSTNAME=$NEW_MASTER_HOSTNAME
TARGET_ARCHIVE=/db/node0archive
STANDBY1_HOSTNAME_TO_RETARGET=$NODE1_HOSTNAME
else
TARGET_HOSTNAME=$NEW_MASTER_HOSTNAME
TARGET_ARCHIVE=/db/node1archive
STANDBY1_HOSTNAME_TO_RETARGET=$NODE0_HOSTNAME
fi;
fi;
ssh -T postgres@$NEW_MASTER_HOSTNAME $PGBIN/pg_ctl promote -D $PGDATA
ssh -T postgres@$STANDBY1_HOSTNAME_TO_RETARGET $LOCALBIN/retargetMaster.sh $TARGET_HOSTNAME $TARGET_ARCHIVE
exit 0;
else
TARGET_HOSTNAME=0.0.0.0
TARGET_ARCHIVE=/db/archive
NEW_STANDBY_HOSTNAME=0.0.0.0
if [ $OLD_PRIMARY_NODEID = 0 ]; then
TARGET_HOSTNAME=$NODE0_HOSTNAME
TARGET_ARCHIVE=$NODE0_ARCHIVE
if [ $FAILED_NODEID = 1 ]; then
NEW_STANDBY_HOSTNAME=$NODE2_HOSTNAME
else
NEW_STANDBY_HOSTNAME=$NODE1_HOSTNAME
fi
fi;
if [ $OLD_PRIMARY_NODEID = 1 ]; then
TARGET_HOSTNAME=$NODE1_HOSTNAME
TARGET_ARCHIVE=$NODE1_ARCHIVE
if [ $FAILED_NODEID = 2 ]; then
NEW_STANDBY_HOSTNAME=$NODE0_HOSTNAME
else
NEW_STANDBY_HOSTNAME=$NODE2_HOSTNAME
fi
fi;
if [ $OLD_PRIMARY_NODEID = 2 ]; then
TARGET_HOSTNAME=$NODE2_HOSTNAME
TARGET_ARCHIVE=$NODE2_ARCHIVE
if [ $FAILED_NODEID = 0 ]; then
NEW_STANDBY_HOSTNAME=$NODE1_HOSTNAME
else
NEW_STANDBY_HOSTNAME=$NODE0_HOSTNAME
fi
fi;
ssh -T postgres@$NEW_STANDBY_HOSTNAME $LOCALBIN/retargetMaster.sh $TARGET_HOSTNAME $TARGET_ARCHIVE
exit 0;
fi;
echo 'Did not complete failover'
echo -------------------End failover---------------
exit 0;
echo ------------------------Begin failover------------------------
#------------------------------------
# Inputs
#------------------------------------
FAILED_NODEID=$1 # %d = node id
FAILED_HOSTNAME=$2 # %h = hostname
FAILED_DBPATH=$3 # %D = database cluster path
NEW_MASTER_NODEID=$4
NEW_MASTER_HOSTNAME=$5
NEW_MASTER_DBPATH=$6
OLD_MASTER_NODEID=$7
OLD_PRIMARY_NODEID=$8
# %m = new master node id
# %H = new master host name
# %R = new master database cluster path
# %M = old master node id
# %P = old primary node id
#------------------------
# Nodes
#------------------------
PGDATA=/db/postgres
PGBIN=/usr/pgsql-10.x/bin
LOCALBIN=/db/bin
NODE0_HOSTNAME=10.91.9.200
NODE0_ARCHIVE=/db/node0archive
NODE1_HOSTNAME=10.91.9.24
NODE1_ARCHIVE=/db/node1archive
NODE2_HOSTNAME=10.91.9.41
NODE2_ARCHIVE=/db/node2archive
if [ $FAILED_NODEID = $OLD_PRIMARY_NODEID ]; then
if ssh -T postgres@$NEW_MASTER_HOSTNAME grep -q "$FAILED_HOSTNAME" /db/postgres/recovery.conf; then
#yes, it is, go ahead to promote
echo Node $NEW_MASTER_HOSTNAME can be promoted safely
else
#otherwise, new master has to be switched.
if [ $FAILED_NODEID = 1 ]; then
if [ $NEW_MASTER_NODEID = 0 ]; then
NEW_MASTER_HOSTNAME=$NODE2_HOSTNAME
NEW_MASTER_NODEID=2
else
NEW_MASTER_HOSTNAME=$NODE0_HOSTNAME
NEW_MASTER_NODEID=0
fi;
fi;
if [ $FAILED_NODEID = 2 ]; then
if [ $NEW_MASTER_NODEID = 0 ]; then
NEW_MASTER_HOSTNAME=$NODE1_HOSTNAME
NEW_MASTER_NODEID=1
else
NEW_MASTER_HOSTNAME=$NODE0_HOSTNAME
NEW_MASTER_NODEID=0
fi;
fi;
if [ $FAILED_NODEID = 0 ]; then
if [ $NEW_MASTER_NODEID = 1 ]; then
NEW_MASTER_HOSTNAME=$NODE2_HOSTNAME
NEW_MASTER_NODEID=2
else
NEW_MASTER_HOSTNAME=$NODE1_HOSTNAME
NEW_MASTER_NODEID=1
fi;
fi;
echo New master has been switched to $NEW_MASTER_HOSTNAME
fi;
TARGET_HOSTNAME=0.0.0.0
TARGET_ARCHIVE=/db/archive
STANDBY1_HOSTNAME_TO_RETARGET=0.0.0.0
if [ $FAILED_NODEID = 0 ]; then
if [ $NEW_MASTER_NODEID = 1 ]; then
TARGET_HOSTNAME=$NEW_MASTER_HOSTNAME
TARGET_ARCHIVE=/db/node1archive
STANDBY1_HOSTNAME_TO_RETARGET=$NODE2_HOSTNAME
else
TARGET_HOSTNAME=$NEW_MASTER_HOSTNAME
TARGET_ARCHIVE=/db/node2archive
STANDBY1_HOSTNAME_TO_RETARGET=$NODE1_HOSTNAME
fi;
fi;
if [ $FAILED_NODEID = 1 ]; then
if [ $NEW_MASTER_NODEID = 0 ]; then
TARGET_HOSTNAME=$NEW_MASTER_HOSTNAME
TARGET_ARCHIVE=/db/node0archive
STANDBY1_HOSTNAME_TO_RETARGET=$NODE2_HOSTNAME
else
TARGET_HOSTNAME=$NEW_MASTER_HOSTNAME
TARGET_ARCHIVE=/db/node2archive
STANDBY1_HOSTNAME_TO_RETARGET=$NODE0_HOSTNAME
fi;
fi;
if [ $FAILED_NODEID = 2 ]; then
if [ $NEW_MASTER_NODEID = 0 ]; then
TARGET_HOSTNAME=$NEW_MASTER_HOSTNAME
TARGET_ARCHIVE=/db/node0archive
STANDBY1_HOSTNAME_TO_RETARGET=$NODE1_HOSTNAME
else
TARGET_HOSTNAME=$NEW_MASTER_HOSTNAME
TARGET_ARCHIVE=/db/node1archive
STANDBY1_HOSTNAME_TO_RETARGET=$NODE0_HOSTNAME
fi;
fi;
ssh -T postgres@$NEW_MASTER_HOSTNAME $PGBIN/pg_ctl promote -D $PGDATA
ssh -T postgres@$STANDBY1_HOSTNAME_TO_RETARGET $LOCALBIN/retargetMaster.sh $TARGET_HOSTNAME $TARGET_ARCHIVE
exit 0;
else
TARGET_HOSTNAME=0.0.0.0
TARGET_ARCHIVE=/db/archive
NEW_STANDBY_HOSTNAME=0.0.0.0
if [ $OLD_PRIMARY_NODEID = 0 ]; then
TARGET_HOSTNAME=$NODE0_HOSTNAME
TARGET_ARCHIVE=$NODE0_ARCHIVE
if [ $FAILED_NODEID = 1 ]; then
NEW_STANDBY_HOSTNAME=$NODE2_HOSTNAME
else
NEW_STANDBY_HOSTNAME=$NODE1_HOSTNAME
fi
fi;
if [ $OLD_PRIMARY_NODEID = 1 ]; then
TARGET_HOSTNAME=$NODE1_HOSTNAME
TARGET_ARCHIVE=$NODE1_ARCHIVE
if [ $FAILED_NODEID = 2 ]; then
NEW_STANDBY_HOSTNAME=$NODE0_HOSTNAME
else
NEW_STANDBY_HOSTNAME=$NODE2_HOSTNAME
fi
fi;
if [ $OLD_PRIMARY_NODEID = 2 ]; then
TARGET_HOSTNAME=$NODE2_HOSTNAME
TARGET_ARCHIVE=$NODE2_ARCHIVE
if [ $FAILED_NODEID = 0 ]; then
NEW_STANDBY_HOSTNAME=$NODE1_HOSTNAME
else
NEW_STANDBY_HOSTNAME=$NODE0_HOSTNAME
fi
fi;
ssh -T postgres@$NEW_STANDBY_HOSTNAME $LOCALBIN/retargetMaster.sh $TARGET_HOSTNAME $TARGET_ARCHIVE
exit 0;
fi;
echo 'Did not complete failover'
echo -------------------End failover---------------
exit 0;
Rendere lo script eseguibile:
chmod a+x /etc/pgpool-II/failover.sh
Passo 3
Aggiornare pcp.conf.
vim /etc/pgpool-II/pcp.conf
con il contenuto seguente:
# Lines beginning with '#' (pound) are comments and will
# be ignored. Again, no spaces or tabs allowed before '#'.postgres:e8a48653851e28c69d0506508fb27fc5
# USERID:MD5PASSWD
# be ignored. Again, no spaces or tabs allowed before '#'.postgres:e8a48653851e28c69d0506508fb27fc5
# USERID:MD5PASSWD
La password di default per PostgreSQL è "postgres", pertanto è necessario aggiornarla.
Passo 4
Aggiornare pool_hba.conf.
vim /etc/pgpool-II/pool_hba.conf
con il contenuto seguente:
# "local" is for Unix domain socket connections only
local all all trust
# IPv4 local connections:
host all all 10.91.0.0/16 trust
local all all trust
# IPv4 local connections:
host all all 10.91.0.0/16 trust
È possibile modificarlo in base ai requisiti di protezione.