场景7 Data Guard
官方文档 :Oracle Data Guard Concepts and Administration
用于数据容灾,通过主备库同步(主库将redo日志传送到备库,一个主库可对应多个备库(30)),实现异地容灾
1. 构建测试库
2. 构建读写分离的环境
3. 实现滚动升级
4. 数据备份
数据容灾 :
1. 基于存储复制
2. 基于逻辑卷复制(镜像)
3. 基于应用(redo log)
redo log : DML/DDL, 记录数据块变化,用于recovery
RAC 高可用性,一个库可用多个实例来访问
Golden Gate : 更灵活,成本更高
物理备库 :收到redo日志后,直接做recovery
逻辑备库 :收到redo日志后,通过logminer抽取日志中sql, 执行sql
1. physical dg : 相当于主库的克隆,相同的名字,相同的id,支持主库所有的应用,在open状态下只读 read only(Active)可将所有写的操作放在主库上,所有读的操作放在备库上
2. logical dg : 和主库是两个独立的数据库,不同的名字,不同的id, 可以open到读写read write状态,(部分DDL操作及数据类型不受支持,如 lob, 一般可用于建立物化视图,建立辅助索引,升级
1. physical dg的构建
2. 保护模式的切换(最大保护,最佳性能(默认),最高可用性)
3. snapshot dg
4. dg 切换
5. broker (FSF) 自动切换
最大保护模式maximum protection(同步方式sync传送日志) :备库RFS确认后(ack),主库才能完成提交任务,若主库一直收不到确认(默认10分钟),则自动shutdown abort, 因此该模式需要至少两个以上的备库,可以保证数据零丢失(主备库间无任何数据差异),但对生产环境影响 较大,网络必须快,备库不能有问题,不然影响主库
LNS (log network service) : 主库发送日志
RFS (receive file service) :备库接收日志,写入本地standby redo log files
最佳性能模式maximum performance(异步方式async传送日志):不需备库确认日志收到后,主库就能完成提交,若redo日志传送失败,且主库已发生日志切换,可讲archive log传送给备库
最高可用性模式maximum availability(同步方式传送日志):备库RFS确认后(ack),主库才能完成提交任务(最大保护模式),若主库一直收不到确认(默认10分钟),则自动转换成最佳性能模式,恢复正常后,又自动转换成最大保护模式
MRP (media recover process) : 介质恢复进程
LSP (logical standby coordinator process) : 逻辑
eg :
desc v$database
select name, force_logging from v$database;
alter database force logging;
archive log list;
搭建DG :
1. 修改初始化文件
eg :
show parameter spfile;
create pfile from spfile;
cd $oracle_home/db
vi init.ora
GAP : 日志间隙
FAL_SERVER :
DB_FILE_NAME_CONVERT=‘/u01/app/oracle/oradata/shdb/‘, ‘/u01/app/oracle/oradata/prod/‘, ‘/u01/app/oracle/oradata/shdb/‘, ‘/u02/app/oracle/oradata/prod/‘
LOG)FILE_NAME_CONVERT=…
数据文件路径转换 :主备库数据文件路径不一致
eg :
select name from v$datafile;
select member from v4logfile;
eg :
startup force gnomon pfile=‘ORACLE_HOME/dbs/initprod.ora’;
show parameter name
show parameter log
create spoil from file;
startup force mount;
alter database create standby controlfile as ‘/home/oracle/std_control01.ctl’;
su -oracle
tail -f /u01/app/oracle/diag/
vi etc/hosts
…
ping …
cd $ORACLE_HOME/dbs
ls
scp initprod.ora enmo:$ORACLE_HOME/dbs/initshdb.ora
scp orapwprod enmo:$ORACLE_HOME/dbs/orapwshdb
scp /home/oracle/std_control
cd /u01/app/oracle/oradata/prod
scp *.dbf enmo:/u01/app/oracle/oradata/shdb
备库:
mkdir -p /u01/app/oracle/oradata/shdb
chown -R oracle:oinstall /u01/app/oracle/oradata/shdb
mkdir op /dsk1/arch_shdb
chown -R oracle:oinstall /dsk/arch_shdb
recover managed standby database disconnect from session;
recover managed standby database cancel;
主备库切换 :
1. switchover 正常手工切换
2. failover 容灾切换 :主库down了,将备库强制切换成主库
eg :
alter system switch logfile;
select name, dbid, database_role, protection_mode, switchover_status from v$database;
select username, sid from v$session where username is not null;
alter database commit to switchover to standby; (主库切换到备库)
alter database commit to switchover to standby with session shutdown; (主库切换到备库, 强制关闭会话)
shutdown immediate;
startup mount;
select name, dbid, database_role, protection_mode, switchover_status from v$database;
备库 :
select name, dbid, database_role, protection_mode, switchover_status from v$database;
select username, sid from v$session where username is not null;
select name, dbid, database_role, protection_mode, switchover_status from v$database;
alter system kill session; (强制杀掉绘画)
alter database commit to switchover to primary;
alter database open;
recover managed standby database disconnect from session;
select max(sequence#) from v$archived_log;
desc v$archive_dest
select name, database_role, protection_mode, switchover_status from v$database;
snapshot dg : 用于应用测试,建立快照,生成一个还原点,测试期间不能和主库同步,但可以继续接收主库的日志,测试完成后,回到还原点,开始recover
snapshot 保存到 recover area
show parameter recover
alter system set db_recovery_file_dest_size = 2g;
alter system set db_recovery_file_dest = ‘/dsk1’;
alter database convert to snapshot standby;
show parameter recover
show parameter spfile
create spfile from pfile;
startup force mount;
chown -R oracle:dba /dsk1
切换模式 :
shutdown immediate;
startup mount
alter database set standby database to maximize protection;
select name, database_role, protection_mode from v$database;
alter database open;
备库 :
select name, database_role, protection_mode from v$database;
alter database open;
eg :
alter database add standby logfile ‘/u01/app/oracle/oradata/shdb/std_redo01.log’ size 50m;
alter database add standby logfile ‘/u01/app/oracle/oradata/shdb/std_redo02.log’ size 50m;
alter database add standby logfile ‘/u01/app/oracle/oradata/shdb/std_redo03.log’ size 50m;
alter database add standby logfile ‘/u01/app/oracle/oradata/shdb/std_redo04.log’ size 50m;
select member from v$logfile;
select group#, sequence#, status from v$standby_log;
eg :
conn scott/tiger
insert into emp1 select * from emp;
commit
insert into emp1 select * from emp;
select count(*) from emp1;
commit;
主库 :
ifconfig
默认 :MRP只对archive log
eg ;
select group#, sequence#, status from v$standby_log;
recover managed standby database using current logfile disconnect from session; (启动实时应用)
Data Guard Broker
dgmgrl
connect sys/oracle@bjdb
create configuration ‘bjdbcfg’ as primary database is ‘bjdb’ connect identifier is ‘bjdb’;
show configuration
add database ‘shdb’ as connect identifier is shdb maintained as physical;
show configuration
enable configuration; (是Broker生效)
show configuration
edit database ‘bjdb’ set property ‘logxptmode’ = ‘sync’;
edit database ‘shdb’ set property ‘logxptmode’ = ‘sync’;
enable fast_start failover;
show configuration
srvctl (需装GI)