背景
环境
Centos7 minimal
主机规划
server0 192.168.80.100 gtm
server1 192.168.80.101 gtm-proxy1,coordinator1,datanode1
server2 192.168.80.102 gtm-proxy2,coordinator2,datanode2
server3 192.168.80.103 gtm-proxy3,coordinator3,datanode3设置主机名
hostname server0 # 临时修改
hostnamectl set-hostname server0 #
- 修改主机hosts文件
echo "192.168.80.100 server0" >> /etc/hosts
echo "192.168.80.101 server1" >> /etc/hosts
echo "192.168.80.102 server2" >> /etc/hosts
echo "192.168.80.103 server3" >> /etc/hosts
- 添加postgres用户
建立组：
# groupadd postgres
建立用户：
# useradd -rmg postgres postgres
初始化密码：
# passwd postgres
- 配置免密登陆
# root用户
ssh-keygen
ssh-copy-id -i ~/.ssh/id_rsa.pub root@server0
ssh-copy-id -i ~/.ssh/id_rsa.pub root@server1
ssh-copy-id -i ~/.ssh/id_rsa.pub root@server2
ssh-copy-id -i ~/.ssh/id_rsa.pub root@server3
# postgres用户
ssh-keygen
ssh-copy-id -i ~/.ssh/id_rsa.pub postgres@server0
ssh-copy-id -i ~/.ssh/id_rsa.pub postgres@server1
ssh-copy-id -i ~/.ssh/id_rsa.pub postgres@server2
ssh-copy-id -i ~/.ssh/id_rsa.pub postgres@server3
#如果配置完成后，发现ssh并不免密，可按下面操作进行完善：
chmod 755 /home/postgres
chmod 700 /home/postgres/.ssh
chmod 644 /home/postgres/.ssh/authorized_keys
- 安装编译依赖
yum install gcc bison flex perl-ExtUtils-Embed perl python-devel tcl-devel readline-devel openssl-devel krb5-devel e2fsprogs-devel libxml2-devel libxslt-devel pam-devel libuuid-devel openldap-devel openjade opensp docbook-dtds libicu-devel gettext systemd-devel
- 编译安装
#上传源码包至 /data/postgres-xl-10r1.tar.gz
cd /data
tar xf /data/postgres-xl-10r1.tar.gz
mkdir build
cd build
/data/postgres-xl-10r1/configure --prefix=/usr/local/pgsql --with-icu --with-tcl --with-perl --with-python --with-gssapi --with-pam --with-ldap --with-openssl --with-systemd --with-libxml --with-libxslt
make -j4 world
make install-world
- 设置【postgres用户的环境变量】
# vi ~/.bashrc
export PGUSER=postgres
export PGHOME=/usr/local/pgsql
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$PGHOME/lib
export PATH=$PATH:$PGHOME/bin
export TEMP=/tmp
export TMPDIR=/tmp
- 使环境变量生效：
$ source ~/.bashrc
#验证环境变量是否生效：
pg_ctl --version
- 配置Postgres-XL集群
在GTM节点上以postgres用户执行：
$ pgxc_ctl ---初次执行，会提示Error说没有配置文件，忽略即可
PGXC prepare ---执行该命令将会生成一份配置文件模板
PGXC exit --退出 pgxc_ctl交互窗
执行完成后，在postgres用户根目录下，会生成一个pgxc_ctl目录，编辑其中的pgxc_ctl.conf文件
修改为：
#!/usr/bin/env bash
#
# Postgres-XC Configuration file for pgxc_ctl utility.
#
# Configuration file can be specified as -c option from pgxc_ctl command. Default is
# $PGXC_CTL_HOME/pgxc_ctl.org.
#
# This is bash script so you can make any addition for your convenience to configure
# your Postgres-XC cluster.
#
# Please understand that pgxc_ctl provides only a subset of configuration which pgxc_ctl
# provide. Here's several several assumptions/restrictions pgxc_ctl depends on.
#
# 1) All the resources of pgxc nodes has to be owned by the same user. Same user means
# user with the same user name. User ID may be different from server to server.
# This must be specified as a variable $pgxcOwner.
#
# 2) All the servers must be reacheable via ssh without password. It is highly recommended
# to setup key-based authentication among all the servers.
#
# 3) All the databases in coordinator/datanode has at least one same superuser. Pgxc_ctl
# uses this user to connect to coordinators and datanodes. Again, no password should
# be used to connect. You have many options to do this, pg_hba.conf, pg_ident.conf and
# others. Pgxc_ctl provides a way to configure pg_hba.conf but not pg_ident.conf. This
# will be implemented in the later releases.
#
# 4) Gtm master and slave can have different port to listen, while coordinator and datanode
# slave should be assigned the same port number as master.
#
# 5) Port nuber of a coordinator slave must be the same as its master.
#
# 6) Master and slave are connected using synchronous replication. Asynchronous replication
# have slight (almost none) chance to bring total cluster into inconsistent state.
# This chance is very low and may be negligible. Support of asynchronous replication
# may be supported in the later release.
#
# 7) Each coordinator and datanode can have only one slave each. Cascaded replication and
# multiple slave are not supported in the current pgxc_ctl.
#
# 8) Killing nodes may end up with IPC resource leak, such as semafor and shared memory.
# Only listening port (socket) will be cleaned with clean command.
#
# 9) Backup and restore are not supported in pgxc_ctl at present. This is a big task and
# may need considerable resource.
#
#========================================================================================
#
#
# pgxcInstallDir variable is needed if you invoke "deploy" command from pgxc_ctl utility.
# If don't you don't need this variable.
## 配置 pgxcInstallDir 和 pgxlDATA，后续变量引用这两个变量
pgxcInstallDir=$PGHOME
pgxlDATA=/data
#---- OVERALL -----------------------------------------------------------------------------
# 配置 用户名为 postgres
pgxcOwner=postgres # owner of the Postgres-XC databaseo cluster. Here, we use this
# both as linus user and database user. This must be
# the super user of each coordinator and datanode.
pgxcUser=$pgxcOwner # OS user of Postgres-XC owner
tmpDir=/tmp # temporary dir used in XC servers
localTmpDir=$tmpDir # temporary dir used here locally
configBackup=n # If you want config file backup, specify y to this value.
configBackupHost=pgxc-linker # host to backup config file
configBackupDir=$HOME/pgxc # Backup directory
configBackupFile=pgxc_ctl.bak # Backup file name --> Need to synchronize when original changed.
#---- GTM ------------------------------------------------------------------------------------
# GTM is mandatory. You must have at least (and only) one GTM master in your Postgres-XC cluster.
# If GTM crashes and you need to reconfigure it, you can do it by pgxc_update_gtm command to update
# GTM master with others. Of course, we provide pgxc_remove_gtm command to remove it. This command
# will not stop the current GTM. It is up to the operator.
#---- GTM Master -----------------------------------------------
#---- Overall ----
# 配置 GTM
gtmName=gtm
gtmMasterServer=server0
gtmMasterPort=20001
gtmMasterDir=$pgxlDATA/pgxc/nodes/gtm
#---- Configuration ---
gtmExtraConfig=none # Will be added gtm.conf for both Master and Slave (done at initilization only)
gtmMasterSpecificExtraConfig=none # Will be added to Master's gtm.conf (done at initialization only)
#---- GTM Slave -----------------------------------------------
# Because GTM is a key component to maintain database consistency, you may want to configure GTM slave
# for backup.
#---- Overall ------
# 不要 gtmSlave ，所以直接设置 gtmSlave=n
gtmSlave=n # Specify y if you configure GTM Slave. Otherwise, GTM slave will not be configured and
# all the following variables will be reset.
gtmSlaveName=gtmSlave
gtmSlaveServer=node12 # value none means GTM slave is not available. Give none if you don't configure GTM Slave.
gtmSlavePort=20001 # Not used if you don't configure GTM slave.
gtmSlaveDir=$HOME/pgxc/nodes/gtm # Not used if you don't configure GTM slave.
# Please note that when you have GTM failover, then there will be no slave available until you configure the slave
# again. (pgxc_add_gtm_slave function will handle it)
#---- Configuration ----
gtmSlaveSpecificExtraConfig=none # Will be added to Slave's gtm.conf (done at initialization only)
#---- GTM Proxy -------------------------------------------------------------------------------------------------------
# GTM proxy will be selected based upon which server each component runs on.
# When fails over to the slave, the slave inherits its master's gtm proxy. It should be
# reconfigured based upon the new location.
#
# To do so, slave should be restarted. So pg_ctl promote -> (edit postgresql.conf and recovery.conf) -> pg_ctl restart
#
# You don't have to configure GTM Proxy if you dont' configure GTM slave or you are happy if every component connects
# to GTM Master directly. If you configure GTL slave, you must configure GTM proxy too.
#---- Shortcuts ------
gtmProxyDir=$pgxlDATA/pgxc/nodes/gtm_pxy
#---- Overall -------
gtmProxy=y # Specify y if you conifugre at least one GTM proxy. You may not configure gtm proxies
# only when you dont' configure GTM slaves.
# If you specify this value not to y, the following parameters will be set to default empty values.
# If we find there're no valid Proxy server names (means, every servers are specified
# as none), then gtmProxy value will be set to "n" and all the entries will be set to
# empty values.
# 根据需要配置 gtmProxy的个数，建议gtmProxy、coordinator、datanode的一致部署，减少网络开销
gtmProxyNames=(gtm_pxy1 gtm_pxy2 gtm_pxy3) # No used if it is not configured
gtmProxyServers=(server1 server2 server3) # Specify none if you dont' configure it.
gtmProxyPorts=(20001 20001 20001) # Not used if it is not configured.
gtmProxyDirs=($gtmProxyDir $gtmProxyDir $gtmProxyDir) # Not used if it is not configured.
#---- Configuration ----
gtmPxyExtraConfig=none # Extra configuration parameter for gtm_proxy. Coordinator section has an example.
gtmPxySpecificExtraConfig=(none none none)
#---- Coordinators ----------------------------------------------------------------------------------------------------
#---- shortcuts ----------
coordMasterDir=$pgxlDATA/pgxc/nodes/coord
coordSlaveDir=$pgxlDATA/pgxc/nodes/coord_slave
coordArchLogDir=$pgxlDATA/pgxc/nodes/coord_archlog
#---- Overall ------------
coordNames=(coord1 coord2 coord3) # Master and slave use the same name
coordPorts=(20004 20005 20004) # Master ports
poolerPorts=(20010 20011 20010) # Master pooler ports
# 设置所有coordinator 的pg_hba.conf文件
coordPgHbaEntries=(192.168.80.0/24) # Assumes that all the coordinator (master/slave) accepts
# the same connection
# This entry allows only $pgxcOwner to connect.
# If you'd like to setup another connection, you should
# supply these entries through files specified below.
# Note: The above parameter is extracted as "host all all 0.0.0.0/0 trust". If you don't want
# such setups, specify the value () to this variable and suplly what you want using coordExtraPgHba
# and/or coordSpecificExtraPgHba variables.
#coordPgHbaEntries=(::1/128) # Same as above but for IPv6 addresses
#---- Master -------------
coordMasterServers=(server1 server2 server3) # none means this master is not available
coordMasterDirs=($coordMasterDir $coordMasterDir $coordMasterDir)
coordMaxWALsernder=5 # max_wal_senders: needed to configure slave. If zero value is specified,
# it is expected to supply this parameter explicitly by external files
# specified in the following. If you don't configure slaves, leave this value to zero.
coordMaxWALSenders=($coordMaxWALsernder $coordMaxWALsernder $coordMaxWALsernder)
# max_wal_senders configuration for each coordinator.
#---- Slave -------------
# 不要coordSlave，设为n
coordSlave=n # Specify y if you configure at least one coordiantor slave. Otherwise, the following
# configuration parameters will be set to empty values.
# If no effective server names are found (that is, every servers are specified as none),
# then coordSlave value will be set to n and all the following values will be set to
# empty values.
coordUserDefinedBackupSettings=n # Specify whether to update backup/recovery
# settings during standby addition/removal.
coordSlaveSync=y # Specify to connect with synchronized mode.
coordSlaveServers=(node07 node08 node09 node06) # none means this slave is not available
coordSlavePorts=(20004 20005 20004 20005) # Master ports
coordSlavePoolerPorts=(20010 20011 20010 20011) # Master pooler ports
coordSlaveDirs=($coordSlaveDir $coordSlaveDir $coordSlaveDir $coordSlaveDir)
coordArchLogDirs=($coordArchLogDir $coordArchLogDir $coordArchLogDir $coordArchLogDir)
#---- Configuration files---
# Need these when you'd like setup specific non-default configuration
# These files will go to corresponding files for the master.
# You may supply your bash script to setup extra config lines and extra pg_hba.conf entries
# Or you may supply these files manually.
coordExtraConfig=coordExtraConfig # Extra configuration file for coordinators.
# This file will be added to all the coordinators'
# postgresql.conf
# Pleae note that the following sets up minimum parameters which you may want to change.
# You can put your postgresql.conf lines here.
cat > $coordExtraConfig <<EOF
#================================================
# Added to all the coordinator postgresql.conf
# Original: $coordExtraConfig
log_destination = 'stderr'
logging_collector = on
log_directory = 'pg_log'
listen_addresses = '*'
max_connections = 100
EOF
# Additional Configuration file for specific coordinator master.
# You can define each setting by similar means as above.
coordSpecificExtraConfig=(none none none none)
coordExtraPgHba=none # Extra entry for pg_hba.conf. This file will be added to all the coordinators' pg_hba.conf
coordSpecificExtraPgHba=(none none none none)
#----- Additional Slaves -----
#
# Please note that this section is just a suggestion how we extend the configuration for
# multiple and cascaded replication. They're not used in the current version.
#
coordAdditionalSlaves=n # Additional slave can be specified as follows: where you
coordAdditionalSlaveSet=(cad1) # Each specifies set of slaves. This case, two set of slaves are
# configured
cad1_Sync=n # All the slaves at "cad1" are connected with asynchronous mode.
# If not, specify "y"
# The following lines specifies detailed configuration for each
# slave tag, cad1. You can define cad2 similarly.
cad1_Servers=(node08 node09 node06 node07) # Hosts
cad1_dir=$HOME/pgxc/nodes/coord_slave_cad1
cad1_Dirs=($cad1_dir $cad1_dir $cad1_dir $cad1_dir)
cad1_ArchLogDir=$HOME/pgxc/nodes/coord_archlog_cad1
cad1_ArchLogDirs=($cad1_ArchLogDir $cad1_ArchLogDir $cad1_ArchLogDir $cad1_ArchLogDir)
#---- Datanodes -------------------------------------------------------------------------------------------------------
#---- Shortcuts --------------
datanodeMasterDir=$pgxlDATA/pgxc/nodes/dn_master
datanodeSlaveDir=$pgxlDATA/pgxc/nodes/dn_slave
datanodeArchLogDir=$pgxlDATA/pgxc/nodes/datanode_archlog
#---- Overall ---------------
#primaryDatanode=datanode1 # Primary Node.
# At present, xc has a priblem to issue ALTER NODE against the primay node. Until it is fixed, the test will be done
# without this feature.
primaryDatanode=datanode1 # Primary Node 主数据节点 .
datanodeNames=(datanode1 datanode2 datanode3)
datanodePorts=(20008 20009 20008) # Master ports
datanodePoolerPorts=(20012 20013 20012) # Master pooler ports
datanodePgHbaEntries=(192.168.80.0/24) # Assumes that all the coordinator (master/slave) accepts
# the same connection
# This list sets up pg_hba.conf for $pgxcOwner user.
# If you'd like to setup other entries, supply them
# through extra configuration files specified below.
# Note: The above parameter is extracted as "host all all 0.0.0.0/0 trust". If you don't want
# such setups, specify the value () to this variable and suplly what you want using datanodeExtraPgHba
# and/or datanodeSpecificExtraPgHba variables.
#datanodePgHbaEntries=(::1/128) # Same as above but for IPv6 addresses
#---- Master ----------------
datanodeMasterServers=(server1 server2 server3) # none means this master is not available.
# This means that there should be the master but is down.
# The cluster is not operational until the master is
# recovered and ready to run.
datanodeMasterDirs=($datanodeMasterDir $datanodeMasterDir $datanodeMasterDir)
datanodeMaxWalSender=5 # max_wal_senders: needed to configure slave. If zero value is
# specified, it is expected this parameter is explicitly supplied
# by external configuration files.
# If you don't configure slaves, leave this value zero.
datanodeMaxWALSenders=($datanodeMaxWalSender $datanodeMaxWalSender $datanodeMaxWalSender)
# max_wal_senders configuration for each datanode
#---- Slave -----------------
# 不用datanodeSlave ，设置为n
datanodeSlave=n # Specify y if you configure at least one coordiantor slave. Otherwise, the following
# configuration parameters will be set to empty values.
# If no effective server names are found (that is, every servers are specified as none),
# then datanodeSlave value will be set to n and all the following values will be set to
# empty values.
datanodeUserDefinedBackupSettings=n # Specify whether to update backup/recovery
# settings during standby addition/removal.
datanodeSlaveServers=(node07 node08 node09 node06) # value none means this slave is not available
datanodeSlavePorts=(20008 20009 20008 20009) # value none means this slave is not available
datanodeSlavePoolerPorts=(20012 20013 20012 20013) # value none means this slave is not available
datanodeSlaveSync=y # If datanode slave is connected in synchronized mode
datanodeSlaveDirs=($datanodeSlaveDir $datanodeSlaveDir $datanodeSlaveDir $datanodeSlaveDir)
datanodeArchLogDirs=( $datanodeArchLogDir $datanodeArchLogDir $datanodeArchLogDir $datanodeArchLogDir )
# ---- Configuration files ---
# You may supply your bash script to setup extra config lines and extra pg_hba.conf entries here.
# These files will go to corresponding files for the master.
# Or you may supply these files manually.
datanodeExtraConfig=none # Extra configuration file for datanodes. This file will be added to all the
# datanodes' postgresql.conf
datanodeSpecificExtraConfig=(none none none none)
datanodeExtraPgHba=none # Extra entry for pg_hba.conf. This file will be added to all the datanodes' postgresql.conf
datanodeSpecificExtraPgHba=(none none none none)
#----- Additional Slaves -----
datanodeAdditionalSlaves=n # Additional slave can be specified as follows: where you
# datanodeAdditionalSlaveSet=(dad1 dad2) # Each specifies set of slaves. This case, two set of slaves are
# configured
# dad1_Sync=n # All the slaves at "cad1" are connected with asynchronous mode.
# If not, specify "y"
# The following lines specifies detailed configuration for each
# slave tag, cad1. You can define cad2 similarly.
# dad1_Servers=(node08 node09 node06 node07) # Hosts
# dad1_dir=$HOME/pgxc/nodes/coord_slave_cad1
# dad1_Dirs=($cad1_dir $cad1_dir $cad1_dir $cad1_dir)
# dad1_ArchLogDir=$HOME/pgxc/nodes/coord_archlog_cad1
# dad1_ArchLogDirs=($cad1_ArchLogDir $cad1_ArchLogDir $cad1_ArchLogDir $cad1_ArchLogDir)
#---- WAL archives -------------------------------------------------------------------------------------------------
walArchive=n # If you'd like to configure WAL archive, edit this section.
# Pgxc_ctl assumes that if you configure WAL archive, you configure it
# for all the coordinators and datanodes.
# Default is "no". Please specify "y" here to turn it on.
#
# End of Configuration Section
#
#==========================================================================================================================
#========================================================================================================================
# The following is for extension. Just demonstrate how to write such extension. There's no code
# which takes care of them so please ignore the following lines. They are simply ignored by pgxc_ctl.
# No side effects.
#=============<< Beginning of future extension demonistration >> ========================================================
# You can setup more than one backup set for various purposes, such as disaster recovery.
walArchiveSet=(war1 war2)
war1_source=(master) # you can specify master, slave or ano other additional slaves as a source of WAL archive.
# Default is the master
wal1_source=(slave)
wal1_source=(additiona_coordinator_slave_set additional_datanode_slave_set)
war1_host=node10 # All the nodes are backed up at the same host for a given archive set
war1_backupdir=$HOME/pgxc/backup_war1
wal2_source=(master)
war2_host=node11
war2_backupdir=$HOME/pgxc/backup_war2
#=============<< End of future extension demonistration >> ========================================================
- 初始化集群
pgxc_ctl -c ~/pgxc_ctl/pgxc_ctl.conf init all
- 启动集群
pgxc_ctl -c ~/pgxc_ctl/pgxc_ctl.conf start all
- 关闭集群
pgxc_ctl -c ~/pgxc_ctl/pgxc_ctl.conf stop all
- 删除集群
ssh root@server0 "rm -rf /data/pgxc"
ssh root@server1 "rm -rf /data/pgxc"
ssh root@server2 "rm -rf /data/pgxc"
ssh root@server3 "rm -rf /data/pgxc"
