Dameng real-time active and standby cluster construction

Read write separation cluster deployment ​

Hosting plan ​

Host name ​	service ip​	heartbeat ip​	Database name ​	Instance name ​
dm-16​	192.168.3.16​		No instance required , Only for monitoring and management ​	No instance required , Only for monitoring and management ​
dm-17​	192.168.2.17​	192.168.3.17​	dmrw​	dmrw1​
dm-18​	192.168.3.18​	192.168.3.18​	dmrw​	dmrw2​

Port planning ​

Instance name ​	Instance port ​	MAL System monitoring TCP Port of connection ​	The instance local daemon listens TCP Port of connection ​	Instance listening daemon TCP Port of connection ​
dmrw1​	5236​	7336​	7436​	7536​
dmrw2​	5236​	7336​	7436​	7536​

The local store LVM disk ​

disk ​	LVM Disk name / purpose ​
/dev/sda(25G)​	/dev/mapper/centos-root​ Database installation disk )​
/dev/sda(25G)​	/dev/mapper/centos-root​ Data disk )​

Directory planning ​

Database software installation directory ​	/home/dmdba/dmdbms​
Instance installation directory ​	/home/dmdba/dmdata/​
Archive log storage directory ​	/home/dmdba/dmdata/arch​
Backup file storage directory ​	/home/dmdba/dmdata/dmbak/​

Steps to build ​

Modify hostname ​

vi /etc/hostname ​

Modify the host name on the plan ​

Software installation ​

dm Database single node installation steps , Here slightly ​

Initialize instance ​

Both nodes need to operate （dm-17\dm-18）​

dminit path=/home/dmdba/dmdata page_size=32 extent_size=32 charset=1 log_size=2048 db_name=dmrw instance_name=dmrw​

To host dm-17 Separate the main library for reading and writing , Boot instance ​

dmserver /home/dmdba/dmdata/dmrw/dm.ini​

appear system is ready Post input exit Stop database ​

Offline backup database ​

Host computer dmdb01 Execute the following command on , Confirm master library dmap Service started ：​

ps -ef|grep dmap​

dmdba user start-up dmrman Tools ​

dmrman use_ap=2​

perform backup Whole database ：​

backup database '/home/dmdba/dmdata/dmrw/dm.ini' backupset '/home/dmdba/dmdata/dmbak/bakfull';​

Backup restore backup database ​

Backed up files /home/dmdba/dmdba/dmbak/bakfull Copy to server dm-18 On .​

scp -r /home/dmdba/dmdata/dmbak/bakfull 192.168.2.18:/home/dmdba/dmdata/dmbak/.​

Host computer dm-18 On ​

Use dmrman Restore the standby database with tools ,dmdba User execution ：​

dmrman use_ap=2​

Get into dmrman Tool interface ​

perform restore：​

restore database '/home/dmdba/dmdata/dmrw/dm.ini' from backupset '/home/dmdba/dmdata/dmbak/bakfull'​

When it's done recover：​

recover database '/home/dmdba/dmdata/dmrw/dm.ini' from backupset '/home/dmdba/dmdata/dmbak/bakfull '​

Finally, execute recover update db_magic.​

recover database '/home/dmdba/dmdata/dmrw/dm.ini' update db_magic;​

modify dm.ini Parameters ​

All nodes need to be modified ​

Execute the following command ：​

vi /home/dmdba/dmdata/dmrw/dm.ini​

Main library dm-17 Modify the following parameter values on ：​

INSTANCE_NAME = dmrw1​

MAL_INI = 1​

ARCH_INI = 1​

ALTER_MODE_STATUS = 0​

ENABLE_OFFLINE_TS = 2​

For the library dm-18 Modify the following parameter values on ：​

INSTANCE_NAME = dmrw2​

MAL_INI = 1​

ARCH_INI = 1​

ALTER_MODE_STATUS = 0​

ENABLE_OFFLINE_TS = 2​

Configure archive configuration file ​

Create a new file in the instance Directory dmarch.ini, Execute the following command ：​

vi /dmdata/dmdb/dmarch.ini​

Create a new file in the instance Directory dmarch.ini, Execute the following command ：​

vi /dmdata/dmdb/dmarch.ini​

Main library dm-17 Add the following on ：​

[ARCHIVE_REALTIME1] ​

ARCH_TYPE = REALTIME ​

ARCH_DEST = dmrw2 ​

[ARCHIVE_REALTIME2] ​

ARCH_TYPE = REALTIME ​

ARCH_DEST = dmrw3 ​

[ARCHIVE_LOCAL1]​

ARCH_TYPE = LOCAL ​

ARCH_DEST = /home/dmdba/dmdata/arch ​

ARCH_FILE_SIZE = 2048 ​

ARCH_SPACE_LIMIT = 102400​

For the library dm-18 Add the following on ：​

[ARCHIVE_REALTIME1] ​

ARCH_TYPE = REALTIME ​

ARCH_DEST = dmrw1 ​

[ARCHIVE_REALTIME2] ​

ARCH_TYPE = REALTIME ​

ARCH_DEST = dmrw3 ​

[ARCHIVE_LOCAL1]​

ARCH_TYPE = LOCAL ​

ARCH_DEST = /home/dmdba/dmdata/arch ​

ARCH_FILE_SIZE = 2048 ​

ARCH_SPACE_LIMIT = 102400​

To configure dmmal.ini（ Heartbeat Network ）​

Both nodes should be configured , And the configuration must be the same ​

[[email protected] dmrw]$ cat dmmal.ini ​

MAL_CHECK_INTERVAL = 5 ​

MAL_CONN_FAIL_INTERVAL = 15 ​

[MAL_INST1]​

MAL_INST_NAME = dmrw1 ​

MAL_HOST = 192.168.3.17 ​

MAL_PORT = 7336​

MAL_INST_HOST = 192.168.2.17 ​

MAL_INST_PORT = 5236 ​

MAL_DW_PORT = 7436 ​

MAL_INST_DW_PORT = 7536​

[MAL_INST2]​

MAL_INST_NAME = dmrw2 ​

MAL_HOST = 192.168.3.18 ​

MAL_PORT = 7336​

MAL_INST_HOST = 192.168.2.18 ​

MAL_INST_PORT = 5236 ​

MAL_DW_PORT = 7436 ​

MAL_INST_DW_PORT = 7536​

Configuration daemon ​

Both nodes should be configured , And the configuration must be the same ​

[[email protected] dmrw]$ cat dmwatcher.ini ​

[GRP_RW] ​

DW_TYPE = GLOBAL ​

DW_MODE =MANUAL ​

DW_ERROR_TIME = 30 ​

INST_RECOVER_TIME = 60 ​

INST_ERROR_TIME = 20 ​

INST_OGUID = 453331 ​

INST_INI = /home/dmdba/dmdata/dmrw/dm.ini ​

INST_AUTO_RESTART = 1 ​

INST_STARTUP_CMD = /home/dmdba/dmdbms/bin/dmserver​

Start the database instance ​

Use dmdba user , Execute the following command on the main library （ Both active and standby databases execute ）：​

dmserver /home/dmdba/dmdata/dmrw/dm.ini mount​

Use in new terminals disql Tools connect to the database ：​

./disql SYSDBA/SYSDBA​

Both master and slave databases are modified oguid, Execute the following command ：​

sp_set_oguid(453331);​

Main library dm-17 Modify the database mode to primary, Execute the following command ：​

alter database primary;​

For the library dm-18 Similar operations on , But note that the database schema is modified to standby, Execute the following command ：​

alter database standby;​

dmdba Under the user , To the database installation directory bin Execute the following command （ Both active and standby databases execute ）：​

dmwatcher /home/dmdba/dmdata/dmrw/dmwatcher.ini​

Check the status ​

Use DM Management tools , Connect two instances to view ​

Configure the monitor （ The server 192.168.3.16）​

192.168.3.16 First, install the database software, and then operate the following steps ​

vi /home/dmdba/dmdbms/bin/dmmonitor.ini​

Add the following ：​

explain ： Specific configuration , Please put # and # Later Chinese content is deleted .​

MON_DW_CONFIRM = 1 ​

MON_LOG_PATH = /home/dmdba/dmdbms/log ​

MON_LOG_INTERVAL = 60 ​

MON_LOG_FILE_SIZE = 200​

MON_LOG_SPACE_LIMIT = 1024 ​

[GRP_RW] ​

MON_INST_OGUID = 453331 ​

MON_DW_IP = 192.168.3.17:7436​

MON_DW_IP = 192.168.3.18:7436​

Start the monitor ​

On the server dm-16 On the operation ：​

dmmonitor /dmdata/dmdb/dmmonitor.ini​

You can see the cluster status ok​

The above proves that the cluster is built successfully ​

Guanku ​

Close by dmmonitor---dmwatcher After closing （dmwatcher The shutdown sequence starts with the main backup ）---dmserver（dmserver The shutdown sequence starts with the main backup ） Close the database in order ​

Add service as system startup service ​

The following operations should be performed on both the primary and standby nodes , Execution order ： Execute the main database first ---- Standby database execution ：​

cd /home/dmdba/dmdbms/script/root/​

./dm_service_installer.sh -t dmwatcher -p dmrw -watcher_ini /home/dmdba/dmdata/dmrw/dmwatcher.ini​

./dm_service_installer.sh -t dmserver -p dmrw -dm_ini /home/dmdba/dmdata/dmrw/dm.ini​

The monitor executes （192.168.3.16） perform ​

Use root user , To the database installation directory script/root Next .​

cd /home/dmdba/dmdbms/script/root/​

./dm_service_installer.sh -t dmmonitor -p monitor-confirm -monitor_ini /home/dmdba/dmdbms/bin/dmmonitor.ini​

Start command ​

systemctl start DmMonitorServicemonitor-confirm​

host dm-16 Add another ordinary monitor ​

cd /home/dmdba/dmdbms/bin​

cp dmmonitor.ini dmmonitor-normal.ini​

modify dmmonitor-normal.ini​

[[email protected] bin]$ cat dmmonitor-normal.ini ​

MON_DW_CONFIRM = 0 ​

MON_LOG_PATH = /home/dmdba/dmdbms/log ​

MON_LOG_INTERVAL = 60 ​

MON_LOG_FILE_SIZE = 200​

MON_LOG_SPACE_LIMIT = 1024 ​

[GRP_RW] ​

MON_INST_OGUID = 453331 ​

MON_DW_IP = 192.168.3.17:7436​

MON_DW_IP = 192.168.3.18:7436​

Start cluster ：​

Start the standby database instance ​

192.168.2.18 Boot instance ​

systemctl start DmServicedmrw​

Start the main library instance ​

192.168.2.17​

systemctl start DmServicedmrw​

Start the main library daemon ​

192.168.2.17​

systemctl start DmWatcherServicedmrw​

Start the standby database daemon ​

192.168.2.18​

systemctl start DmWatcherServicedmrw​

Start the monitoring service ​

192.168.3.16​

View the cluster status ​

Law 1 ： Check the log ：​

cat /home/dmdba/dmdbms/log/DmMonitorservicemonitor-confirm.log​

Law two ： Front end view ​

dmmonitor path= /home/dmdba/dmdbms/bin/dmmonitor-normal.ini​

Compare and summarize the separation of active and standby and read-write ​

The essence of the separation of active and standby and read-write is the same , All right. redo There are several solutions produced by different log processing ,​

Main library redo​ Standby mode ​	Realtime​	Timely​
High performance ​ Arch_wait_apply=0​​	advantage ：​ 1、 The standby database has little impact on the performance of the primary database ​ 2、 Ensure that the standby database receives it as soon as possible redo journal ​	advantage ：​ There are only a lot more logs in the primary database than in the standby database ​ The standby database and the primary database can ensure the consistency of transactions ​
	shortcoming ： Backup data because redo Not repeated in time , There may be delays ​	shortcoming ：​ The standby database data may be delayed , On the one hand, it comes from redo Delay caused by not repeating in time , On the other hand, it comes from the delay caused by the time when the main database writes online logs ​
	Architecture mode ： Real time active standby ​	Architecture mode ： Common active / standby architecture ​
	Applicable scenario ： Business consistency requirements are general , Scenarios that require high availability of the main library ​	Applicable scenario ： The standby database has general requirements for real-time consistency of business , However, there is a requirement that the backup database for disaster recovery cannot lose transactions ​
Transaction consistency ​ Arch_wait_apply=1​	advantage ： On the basis of meeting transaction consistency , It has the lowest impact on the performance of the main library ​	advantage :​ Strong transaction consistency guarantee , The primary and standby databases can be synchronized ​
	shortcoming ：​ 1、 The performance of the primary database is highly affected by the standby database ​ 2、 When the primary database fails, the data of the standby database and the primary database may be inconsistent ​	shortcoming ：​ The performance of the primary database is affected by the secondary database ,​
	Architecture mode ： Real time read / write separation cluster , The daemon is required to be in automatic mode ​	Architecture pattern : General read-write separation cluster ​
	Applicable scenario ： Business consistency requirements are general , Scenarios that require high availability of the main library ​	Applicable scenario ： High requirements for business consistency , There are no high requirements for the availability of the main database ​

​ Dameng community address ：​​ ​https://eco.dameng.com​​​