达梦数据库REDO误删除恢复疑问？

IT小Chen 2024-05-13

1300

说明:

在没有备份的情况下，经测试非CURRENT REDO LOG FILE丢失可以恢复，CURRENT REDO LOG FILE使用网上的方法无法恢复，初步怀疑新版本对REDO增加了更严格的校验，具体恢复方法还要咨询原厂，测试过程如下。

环境说明:

数据库版本

SQL> SELECT * FROM V$VERSION;


LINEID     BANNER                           
---------- ---------------------------------
1          DM Database Server 64 V8
2          DB Version: 0x7000c
3          03134284172-20240321-222308-20093
4          Msg Version: 0
5          Gsu level(3-4) cnt: 3
6          Gsu level(5) cnt: 0

小版本：

SQL> SELECT BUILD_VERSION FROM V$INSTANCE;


LINEID     BUILD_VERSION                       
---------- ------------------------------------
1          1-3-140-2024.03.21-222308-20093-ENT

补丁版本：

SQL> SELECT ID_CODE;
LINEID     ID_CODE                            
---------- -----------------------------------
1          --03134284172-20240321-222308-20093

安装介质下载地址

https://www.dameng.com/list_103.html

创建测试数据

SQL> CREATE TABLE T1(ID INT);
SQL> INSERT INTO T1 VALUES(1),(2),(3);
SQL> COMMIT;

查询信息

查看REDO信息

SQL> SELECT GROUP_ID,FILE_ID,PATH,RLOG_SIZE FROM V$RLOGFILE;


LINEID     GROUP_ID    FILE_ID     PATH                    RLOG_SIZE           
---------- ----------- ----------- ----------------------- --------------------
1          2           0           dm8/data/CJC/CJC01.log 268435456
2          2           1           dm8/data/CJC/CJC02.log 268435456

SQL> SELECT CKPT_LSN,FILE_LSN,FLUSH_LSN,CUR_LSN,DB_MAGIC FROM V$RLOG;


LINEID     CKPT_LSN             FILE_LSN             FLUSH_LSN            CUR_LSN              DB_MAGIC            
---------- -------------------- -------------------- -------------------- -------------------- --------------------
1          44644                44656                44656                44656                1516351486

新增数据

INSERT INTO T1 VALUES(4),(5);
COMMIT;

SQL> SELECT CKPT_LSN,FILE_LSN,FLUSH_LSN,CUR_LSN,DB_MAGIC FROM V$RLOG; 


LINEID     CKPT_LSN             FILE_LSN             FLUSH_LSN            CUR_LSN              DB_MAGIC            
---------- -------------------- -------------------- -------------------- -------------------- --------------------
1          44644                44661                44661                44661                1516351486

V$ROG和V$RLOGFILE视图说明：

https://eco.dameng.com/document/dm/zh-cn/pm/dm8-admin-manual-appendix#6.%20%E6%97%A5%E5%BF%97%E7%AE%A1%E7%90%86

V$RLOG

显示日志的总体信息。通过该视图可以了解系统当前日志事务号 LSN 的情况、归档日志情况、检查点的执行情况等。

......

V$RLOGFILE

显示日志文件的具体信息。包括文件号、完整路径、文件的状态、文件大小等等。

查看REDO信息

SQL> host ls -lrth dm8/data/CJC/CJC*.log
-rw-r--r-- 1 dmdba dinstall 256M May 11 21:45 dm8/data/CJC/CJC02.log
-rw-r--r-- 1 dmdba dinstall 256M May 11 21:51 dm8/data/CJC/CJC01.log

可以看到，CJC01.log 时间较新，为CURRENT REDO;

模拟故障：

非current redo 丢失

SQL> host mv dm8/data/CJC/CJC02.log dm8/data/CJC/CJC02.log.bak 
SQL> shutdown abort;

启动

[dmdba@cjc-db-01 CJC]$ dmserver dm.ini

前台报错：

/dm8/data/CJC/CJC02.log not exist, can not startup

后台报错：

[dmdba@cjc-db-01 log]$ tail -10f dm_CJC_202405.log
2024-05-11 22:00:21.283 [FATAL] database P0000008747 T0000000000000008747  dm8/data/CJC/CJC02.log not exist,can not startup

启动数据库，找不到CJC02.og，导致启动失败。

用现有的REDO冒充丢失的REDO

[dmdba@cjc-db-01 CJC]$ cp CJC01.log CJC02.log

启动

[dmdba@cjc-db-01 CJC]$ dmserver dm.ini

前台报错

rfil grp init log file dm8/data/CJC/CJC02.log error, code = -717

后台报错

2024-05-11 22:02:32.168 [INFO] database P0000008964 T0000000000000008964  rfil[/dm8/data/CJC/CJC01.log] ver: 7007, sta: 0, arch_lsn = 0, arch_seq = 0, clsn = 44660, clsn_fil = 0, clsn_off = 17547264, l_next_seq = 5058, g_next_seq = 5058, free = 17567744, rpkg_crc_stand = 1
2024-05-11 22:02:32.168 [ERROR] database P0000008964 T0000000000000008964  arch file(/dm8/data/CJC/CJC02.log) fil id check fail
2024-05-11 22:02:32.168 [ERROR] database P0000008964 T0000000000000008964  rfil grp init log file /dm8/data/CJC/CJC02.log error, code = -717
2024-05-11 22:02:32.168 [ERROR] database P0000008964 T0000000000000008964  rlog4_init_low->rfil_grp_alloc failed, alloc_only = 0!
2024-05-11 22:02:32.168 [ERROR] database P0000008964 T0000000000000008964  hlog_upgrade failed, rlog4_sys_init return code[-126]

查看报错：fil id check fail，因为两个REDO fil id不能相同，所以报错。

arch file(/dm8/data/CJC/CJC02.log) fil id check fail

使用dmmdbf工具修改 fil id

dmmdbf工具是什么？

https://eco.dameng.com/document/dm/zh-cn/ops/troubleshooting-human

工具介绍

Dmmdf 工具：主要对 DM 相关的文件的属性进行修改。支持的修改的文件类型有：dbf、rlog、original bak、bakset meta、bakset bkp、bakset，根据文件的不同，可修改的属性也有所不同，最主要的是修改 db_magic。

查看帮助信息

[dmdba@cjc-db-01 CJC]$ dmmdf help
version: 03134284172-20240321-222308-20093
Format:  ./dmmdf KEYWORD=value


Example: ./dmmdf TYPE=1 FILE=/opt/data/DAMENG/SYSTEM.DBF


Keyword            Explanation
--------------------------------------------------------------------------------
TYPE                type
                    1: for dbf
                       bp_flag=0: Modify the normal instance
                       bp_flag=1: Modify the BP instance of DPC
                    2: for rlog
                    3: for original bak
                    4: for bakset meta
                    5: for bakset bkp
                    6: for bakset
                    7: for database degrades
                       scope=1: Only supports from 0x0007000A to 0x00070009, file is dm.ctl path
                       scope=2: Only supports from 0x0007000B to 0x0007000A, file is dm.ini path
                       scope=3: Only supports from 0x0007000C to 0x0007000B for single node database, file is dm.ini path
                    8: for rlog degrades
                       scope=1  Supports from 0x7006 to 0x7005
                       scope=2  Supports from 0x7007 to 0x7006
                    9: for specified database upgrade(Only supports from V8.1.1.88(V7.1.7.88) to V8.1.1.100(V7.1.7.100)
                    10: for updating db_magic of database, file is dm.ini path, db_magic is which update to
                       bp_flag=0: Modify the normal instance
                       bp_flag=1: Modify the BP instance of DPC, also need page_size to be set
FILE                file path
PAGE_SIZE           for TYPE=10 BP_FLAG=1, the page size of the datafile, default 8192
DCR_INI             dmdcr.ini path
SCOPE               scope for database degrades when type=7 or 8
BP_FLAG             whether to modify the BP instance of DPC when type=1 or type=10, default not(0)
DB_MAGIC            target db_magic when type=10
HELP                show this help info
--------------------------------------------------
Example: ./dmmdf TYPE=1 FILE=/opt/data/DAMENG/SYSTEM.dbf
         ./dmmdf TYPE=2 FILE=/opt/data/DAMENG/DAMENG01.log
         ./dmmdf TYPE=3 FILE=/opt/data/DAMENG/bak/dmdb.bak
         ./dmmdf TYPE=4 FILE=/opt/data/DAMENG/bak/bset/nbak.meta
         ./dmmdf TYPE=5 FILE=/opt/data/DAMENG/bak/bset/nbak.bak
         ./dmmdf TYPE=6 FILE=/opt/data/DAMENG/bak/bset
         ./dmmdf TYPE=7 SCOPE=1 FILE=/opt/data/DAMENG/dm.ctl
         ./dmmdf TYPE=7 SCOPE=2 FILE=/opt/data/DAMENG/dm.ini
         ./dmmdf TYPE=7 SCOPE=3 FILE=/opt/data/DAMENG/dm.ini
         ./dmmdf TYPE=8 SCOPE=1 FILE=/opt/data/DAMENG/dm.ini
         ./dmmdf TYPE=8 SCOPE=2 FILE=/opt/data/DAMENG/dm.ini
         ./dmmdf TYPE=9 FILE=/opt/data/DAMENG/dm.ini
         ./dmmdf TYPE=10 FILE=/opt/data/DAMENG/dm.ini DB_MAGIC=123456

修改CJC01.log，将13 fil_id = 0改成13 fil_id = 1

[dmdba@cjc-db-01 CJC]$ dmmdf TYPE=2 FILE=/dm8/data/CJC/CJC01.log
dmmdf V8
**********************************************************
1 sig = DMRLOG
2 ver = 7007
3 chksum = 969006225
4 sta = 0
5 n_magic = 7
6 db_magic = 1516351486
7 len = 268435456
8 free = 17567744
9 clsn = 44660
10 clsn_fil = 0
11 clsn_off = 17547264
12 pemnt_magic = 1510880832
13 fil_id = 0
15 next_seq = 5058
16 g_next_seq = 5058
17 arch_lsn = 0
18 arch_seq = 0
19 dbversion = 0x7000c
20 min_exec_version = V8.1.1.1
21 min_dct_version  = 4
22 p_db_magic = 0
23 n_apply_ep = 0
24 apply_info_lsn = 0
   pkg_seq_arr: (0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0)
   apply_lsn_arr: (0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0)
25 recv_p_db_magic = 0
26 recv_n_apply_ep = 0
   recv_pkg_seq_arr: (0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0)
   recv_apply_lsn_arr: (0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0)
27 l_term_id = 0
28 term_id = 0
29 c_seqno = 5057
30 c_lsn = 44661
31 rpkg_crc_stand = 1
**********************************************************
You can only reset sta(4) or db_magic (6) or clsn (9) or clsn_fil(10) or clsn_off(11) or pemnt_magic(12) or fil_id(13) or next_seq(15) or g_next_seq(16) or p_db_magic(22) or n_apply_ep(23).
Please input the num which one you want to change, q to quit: 13
Input the new value: 1
**********************************************************

再次启动：

[dmdba@cjc-db-01 CJC]$ dmserver dm.ini

启动成功

ndct_db_load_info finished, code:0.
nsvr_process_before_open begin.
nsvr_process_before_open success.
SYSTEM IS READY.

登录数据库，检查

[dmdba@cjc-db-01 CJC]$ disql SYSDBA/CJC123456:5238


Server[LOCALHOST:5238]:mode is normal, state is open
login used time : 15.570(ms)
disql V8


SQL> SELECT * FROM T1;
LINEID     ID         
---------- -----------
1          1
2          2
3          3
4          4
5          5
used time: 1.535(ms). Execute id is 501.


SQL> INSERT INTO T1 VALUES(6);
affect rows 1
used time: 1.171(ms). Execute id is 502.


SQL> COMMIT;
executed successfully
used time: 8.523(ms). Execute id is 503.

current redo 丢失

方法一：失败

按照之前的方法，将另一个redo复制为丢失的redo,使用dmmdf工具只修改fil_id仍然无法启动数据库，

继续修改其他几个值，还是无法启动数据库

sta(4) 
clsn (9)                   44727
clsn_fil(10)
clsn_off(11)               17592320
fil_id(13)                 0
next_seq(15) or            5068
g_next_seq(16) or          5068

方法二：失败

参考https://eco.dameng.com/解决方案，其他机器上重新创建一个相同名称、相同配置的实例，干净关库后考走redo文件到原库，手动修改db_magic等，无法解决问题。

报错如下：

2024-05-11 22:26:31.493 [INFO] database P0000010953 T0000000000000010953  main rfil[/dm8/data/CJC/CJC01.log]'s sta is active
2024-05-11 22:26:31.493 [ERROR] database P0000010953 T0000000000000010953  redo_rfil_grp_pwr_collect->rlog4_first_rpkg_seqno_get failed, code:-723
2024-05-11 22:26:31.494 [INFO] database P0000010953 T0000000000000010953  main rfil [/dm8/data/CJC/CJC01.log]'s grp collect 0 valid pwr record, discard 0 invalid pwr record

初步怀疑新版本对REDO增加了更严格的校验，具体恢复方法还要咨询原厂，有知道解决方法的欢迎留言。

参考链接：

https://eco.dameng.com/community/article/a702397bf5cb0aaf0d018e9cb1bee166
https://eco.dameng.com/community/article/1ffd251e3b2e82b08e0faf0eae9e345c
http://www.killdb.com/2021/08/27/%e8%be%be%e6%a2%a6%e6%95%b0%e6%8d%ae%e5%ba%93%e5%ad%a6%e4%b9%a0%e7%ac%94%e8%ae%b0%e4%b9%8b-current-redo%e8%af%af%e5%88%a0%e9%99%a4%e6%81%a2%e5%a4%8d/

###chenjuchao 20240512###

欢迎关注我的公众号《IT小Chen》

达梦数据库 log

文章转载自IT小Chen，如果涉嫌侵权，请发送邮件至：contact@modb.pro进行举报，并提供相关证据，一经查实，墨天轮将立刻删除相关内容。

达梦数据库REDO误删除恢复疑问？

评论