KingbaseES实现在线增量备份与任意时间点恢复

功能背景

人为操作错误，由于误操作修改或删除了重要数据, 需要KingbaseES 支持还原数据到误操作前的那个时刻

解决方案

提供增量备份能力以及支持任意时间点恢复

实现增量备份和任意时间点恢复的思路是

1. 记录数据库系统的操作记录（WAL归档）
2. 在某一时刻进行一次完整的数据库备份
3. 需要恢复时,还原上一次完整的数据库备份,根据操作记录恢复数据库至指定的某个时刻(即可以实现PITR 时间点恢复)

配置示例

操作
1、配置数据库、vim kingbase.conf

wal_level = replica
archive_mode = on
archive_command = 'cp %p /home/cli/archive/%f && echo %f >> /home/cli/archive/archive.list'
wal_keep_segments=1024

2、重启数据库生效
sys_ctl -D /home/cli/kingbase/db/Server/bin/data/ restart

3、此后数据库的操作都会记录wal日志文件到/home/cli/archive目录下
登录数据库插入数据
create table test(a int primary key);
insert into test values (generate_series(1,50000));
[cli@iZbp1fpui5cmgd2buwhk5fZ bin]$ ls ~/archive/
000000010000000000000001 archive.list

4、创建基础备份，可以使用sys_basebackup或者sys_rman，这里我们用sys_basebackup
sys_basebackup -D /home/cli/kingbase/db/Server/bin/bak -Fp -v -P -Usystem -R

5、测试数据准备

[cli@iZbp1fpui5cmgd2buwhk5fZ bin]$ ./ksql  -Usystem -dtest -p 54321
ksql (V8.0)
Type "help" for help.
test=# create table t2(a int);
CREATE TABLE
test=# insert into t2 values (generate_series(1,1000000));
INSERT 0 1000000
test=# select now();
              now
-------------------------------
 2022-04-23 15:06:03.613826+08
(1 row)

test=#
test=# create table t3(a int);
CREATE TABLE
test=# insert into t3 values (generate_series(1,1000000));
INSERT 0 1000000

6、假定数据库故障，停止数据库，备份故障data并将基础备份的bak替换为新data

sys_ctl -D /home/cli/kingbase/db/Server/bin/data/ stop
mv data data_back
mv bak data

7、修改data目录下kingbase.auto.conf
restore_command = 'cp /home/cli/archive/%f %p > /home/cli/archive/recovery.log 2>&1 ’
recovery_target_time = ‘2022-04-23 15:06:03’

recovery_target_time表示要恢复到的时间点

8、启动数据库
./sys_ctl -D /home/cli/kingbase/db/Server/bin/data/ start

9、查看数据，可知t3未被恢复

日志信息：
2022-04-23 15:17:15.821 CST [2367] LOG: database system was interrupted; last known up at 2022-04-23 14:56:23 CST
2022-04-23 15:17:15.956 CST [2367] LOG: entering standby mode
2022-04-23 15:17:15.975 CST [2367] LOG: restored log file “000000010000000000000003” from archive
2022-04-23 15:17:16.059 CST [2367] LOG: redo starts at 0/3000028
2022-04-23 15:17:16.059 CST [2367] LOG: redo wal segment count 1
2022-04-23 15:17:16.063 CST [2367] LOG: consistent recovery state reached at 0/30000F8
2022-04-23 15:17:16.064 CST [2365] LOG: database system is ready to accept read only connections
2022-04-23 15:17:16.083 CST [2367] LOG: restored log file “000000010000000000000004” from archive
2022-04-23 15:17:16.642 CST [2367] LOG: restored log file “000000010000000000000005” from archive
2022-04-23 15:17:17.239 CST [2367] LOG: restored log file “000000010000000000000006” from archive
2022-04-23 15:17:17.825 CST [2367] LOG: restored log file “000000010000000000000007” from archive
2022-04-23 15:17:18.330 CST [2367] LOG: recovery stopping before commit of transaction 909, time 2022-04-23 15:06:15.005918+08
2022-04-23 15:17:18.330 CST [2367] LOG: recovery has paused
2022-04-23 15:17:18.330 CST [2367] HINT: Execute pg_wal_replay_resume() to continue.

【更多人大金仓数据库信息， 详见 金仓文档管理系统 】