3 #Postgres data directory
4 postgres_datadir='/var/lib/postgresql/9.1/main'
5 #Postgres configuration directory
6 postgres_configdir='/etc/postgresql/9.1/main'
8 postgres_user_key='/var/lib/postgresql/.ssh/id_rsa'
9 #Pgpool configuration directory
10 pgpool_configdir='/etc/pgpool2'
12 if [ -f '/tmp/postgres_master' ]
14 #Get current postgres master id
15 current_master_id=$(cat /tmp/postgres_master);
17 echo "[ERROR] /tmp/postgres_master not found !";
21 #Get postgres master name
22 current_master_name=$(pcp_node_info 10 localhost 9898 postgres postgres $current_master_id | cut -d' ' -f1)
23 #Get postgres slave id
24 [ $current_master_id == 0 ] && current_slave_id=1 || current_slave_id=0
25 #Get postgres slave name
26 current_slave_name=$(pcp_node_info 10 localhost 9898 postgres postgres $current_slave_id | cut -d' ' -f1)
28 #Test if pgpool is running
29 CheckIfPgpoolIsRunning () {
30 #Send signal 0 to pgpool to check if it's running
31 if ! killall -0 pgpool; then echo "[ERROR] Pgpool is not running !"; exit 1; fi;
34 AttachNodeToPgpool () {
35 #pcp_attach_node is a command that permit to attach a specific postgres server (identified by 6th parameter) to pgpool.
36 #pcp_attach_node dont return a good error code when it fails so here if I catch "BackendError" message in stderr I presume
37 #that attachment failed.
38 #TODO:find a condition to break the folowing loop if attachment fails.
39 while [ "`pcp_attach_node 10 localhost 9898 postgres postgres $1`" == "BackendError" ]
41 pcp_attach_node 10 localhost 9898 postgres postgres $1;
42 #This sleep is recommanded to avoid stressing pgpool in this infinite loop.
47 #Whether the slave node is down, start it and attach it to pgpool's backend pool.
48 ReattachDegeneratedSlave () {
50 echo "[INFO] Slave node '$current_slave_name' is down. Performing postgres server reboot..."
51 #Remote postgres reboot via ssh
52 ssh -i $postgres_user_key postgres@$current_slave_name "/etc/init.d/postgresql restart"
53 #Test if postgres is running
54 status=$(ssh -i $postgres_user_key postgres@$current_slave_name "if ! killall -0 postgres; then echo 'error'; else echo 'running'; fi;")
55 if [ $status == "error" ]
57 echo "[ERROR] Postgres slave still down !";
60 echo "[OK] Slave node successfully started.";
63 #Do 'slave online recovery' to force slave sync if it has incoherent data relatevely to master.
64 #echo "[INFO] Starting online recovery for slave '$current_slave_name' ..."
65 #ssh -i /var/lib/postgresql/.ssh/id_rsa postgres@$current_slave_name "bash /var/lib/postgresql/streaming-replication.sh $current_master_name"
66 #Atttach slave (even master) to pgpool's backends pool
67 #Reattach the master node if you have performed an online recovery for slave node and not juste a simple reboot.
68 #Attempting to reatach master to pgpool's backend pool
69 #echo "[INFO] Attaching master node '$current_master_name' ..."
70 #AttachNodeToPgpool "$current_master_id"
71 #echo "[OK] Master node '$current_master_name' has been successfully reattached to pgpool."
72 #Attempting to reattach slave to pgpool's backend pool
73 echo "[INFO] Attaching slave node '$current_slave_name'..."
74 AttachNodeToPgpool "$current_slave_id"
75 echo "[OK] Slave node '$current_slave_name' has been successfully reattached to pgpool."
79 #Whether the master is down do the folowing operations :
80 SwitchOldMasterToSlave () {
82 new_master_name=$current_slave_name
83 new_master_id=$current_slave_id
84 new_slave_name=$current_master_name
85 new_slave_id=$current_master_id
86 #Setup old master config to slave mode
87 echo "[INFO] Setting up configuration for the new slave node '$new_slave_name'..."
88 ssh -i $postgres_user_key postgres@$new_slave_name "/etc/init.d/postgresql stop"
89 ssh -i $postgres_user_key postgres@$new_slave_name "cp -p $postgres_configdir/postgresql.conf.slave $postgres_configdir/postgresql.conf"
90 ssh -i $postgres_user_key postgres@$new_slave_name "[ -f $postgres_datadir/recovery.done ] && mv $postgres_datadir/recovery.done $postgres_datadir/recovery.conf"
91 # Switch slave to new master
92 echo "[INFO] Setting up configuration for the new master '$new_master_name'..."
93 ssh -i $postgres_user_key postgres@$new_master_name "[ -f /tmp/trigger_file ] && rm /tmp/trigger_file"
94 ssh -i $postgres_user_key postgres@$new_master_name "[ -f $postgres_datadir/recovery.conf ] && mv $postgres_datadir/recovery.conf $postgres_datadir/recovery.done"
95 ssh -i $postgres_user_key postgres@$new_master_name "cp -p $postgres_configdir/postgresql.conf.master $postgres_configdir/postgresql.conf"
96 echo "[INFO] Restarting new master..."
97 ssh -i $postgres_user_key postgres@$new_master_name "/etc/init.d/postgresql restart"
98 status=$(ssh -i $postgres_user_key postgres@$new_master_name "if ! killall -0 postgres; then echo 'error'; else echo 'running'; fi;")
99 if [ $status == "error" ]
101 echo "[ERROR] New postgres master not running !";
104 echo "[OK] New master started.";
106 # Start new slave/master with online recovery
107 echo "[INFO] Performing online slave recovery..."
108 ssh -i $postgres_user_key postgres@$new_slave_name "bash /var/lib/postgresql/streaming-replication.sh $new_master_name"
109 echo "[OK] Online recovery completed."
111 #Write changes to pgpool.conf file to keep the same current master and slave nodes even after pgpool reboot.
112 sed -i "s/^backend_hostname0.*/backend_hostname0='$new_master_name'/" $pgpool_configdir/pgpool.conf
113 sed -i "s/^backend_hostname1.*/backend_hostname1='$new_slave_name'/" $pgpool_configdir/pgpool.conf
114 echo "[OK] Pgpool configuration file updated."
116 #Attach new master to pgpool
117 echo "[INFO] Attaching new master node '$new_master_name'..."
118 AttachNodeToPgpool "$new_master_id"
119 echo "[OK] New master node '$new_master_name' has been successfully reattached to pgpool."
121 #Attach new slave to pgpool
122 echo "[INFO] Attaching new slave node '$new_slave_name'..."
123 AttachNodeToPgpool "$new_slave_id"
124 echo "[OK] New slave node '$new_slave_name' has been successfully reattached to pgpool."
128 CheckIfPgpoolIsRunning
130 #Get master/slave state
131 current_master_state=$(pcp_node_info 10 localhost 9898 postgres postgres $current_master_id | cut -d' ' -f3)
132 current_slave_state=$(pcp_node_info 10 localhost 9898 postgres postgres $current_slave_id | cut -d' ' -f3)
134 # state 1 => postgres server is attached but still not receiving connections
135 # state 2 => postgres server is attached and managing clients connections
136 # state 3 => postgres server is detached and probably is down.
138 #If slave is down and master is up then perform an online slave backup.
139 [ $current_slave_state == 3 ] && ([ $current_master_state == 1 ] || [ $current_master_state == 2 ]) && ReattachDegeneratedSlave
140 #If master is down then switch roles between failed master(new server) and the slave(new master).
141 [ $current_master_state == 3 ] && SwitchOldMasterToSlave