分类目录归档:数据库

pg误删除数据恢复(PostgreSQL delete数据恢复)

应用维护人员通过应用程序误删除了一些数据,希望对其进行恢复.通过咨询确认是PostgreSQL数据,wal和arch文件都在,取客户删除数据时间点相关wal文件
QQ20250306-231208


然后让客户提供具体涉及误删除的相关的库和表信息
QQ20250306-230846

通过工具解析wal日志,比较顺利的获取到了undo sql语句

Switch wal to 0000000300000525000000B0 on time 2025-03-05 22:23:14.263661+08
Switch wal to 0000000300000525000000B1 on time 2025-03-05 22:23:14.423082+08
Switch wal to 0000000300000525000000B2 on time 2025-03-05 22:23:15.983833+08
Switch wal to 0000000300000525000000B3 on time 2025-03-05 22:23:17.802107+08
Switch wal to 0000000300000525000000B4 on time 2025-03-05 22:23:18.942125+08
Switch wal to 0000000300000525000000B5 on time 2025-03-05 22:23:20.293585+08
Switch wal to 0000000300000525000000B6 on time 2025-03-05 22:23:21.531484+08
Switch wal to 0000000300000525000000B7 on time 2025-03-05 22:23:24.217501+08
Switch wal to 0000000300000525000000B8 on time 2025-03-05 22:23:27.504164+08
Switch wal to 0000000300000525000000B9 on time 2025-03-05 22:23:29.260754+08
Switch wal to 0000000300000525000000BA on time 2025-03-05 22:23:33.544486+08
Switch wal to 0000000300000525000000BB on time 2025-03-05 22:23:35.568116+08
Switch wal to 0000000300000525000000BC on time 2025-03-05 22:23:37.329659+08
Switch wal to 0000000300000525000000BD on time 2025-03-05 22:23:38.971834+08
Switch wal to 0000000300000525000000BE on time 2025-03-05 22:23:39.922353+08
Switch wal to 0000000300000525000000BF on time 2025-03-05 22:23:40.897294+08

QQ20250306-231657


然后把这些sql语句反向插入到生产库,完成这些误操作数据的恢复
QQ20250306-231935

如果此类的数据库(oracle,mysql,sql server,PostgreSQL)等恢复请求,需要专业恢复技术支持,请联系我们:
电话/微信:17813235971    Q Q:107644445QQ咨询惜分飞    E-Mail:dba@xifenfei.com

发表在 PostgreSQL恢复 | 留下评论

PostgreSQL表文件损坏恢复—pdu恢复损坏的表文件

在某些情况下,由于PostgreSQL表文件损坏导致无法正常访问,可以通过pdu把好的block中的数据恢复出来
准备一张测试表,里面有97条记录

his5_dms=# \d hiscrm.t_sys_oper_log;
                                            Table "hiscrm.t_sys_oper_log"
       Column       |          Type          | Collation | Nullable |Default                     
--------------------+------------------------+-----------+----------+------------
 id                 | bigint                 |           | not null | 
 module             | character varying(50)  |           |          | 
 title              | character varying(50)  |           |          | 
 alias              | character varying(50)  |           |          | 
 business_type      | integer                |           |          | 0
 method             | character varying(200) |           |          | 
 request_method     | character varying(10)  |           |          | 
 operator_type      | integer                |           |          | 0
 oper_name          | character varying(50)  |           |          | 
 dept_name          | character varying(50)  |           |          | 
 oper_url           | character varying(255) |           |          | 
 oper_ip            | character varying(50)  |           |          | 
 oper_location      | character varying(255) |           |          | 
 oper_param         | text                   |           |          | 
 json_result        | text                   |           |          | 
 status             | integer                |           |          | 0
 error_msg          | text                   |           |          | 
 oper_time          | date                   |           |          | 
 create_id          | bigint                 |           |          | 
 create_time        | bigint                 |           |          | 
 clinic_id          | bigint                 |           |          | 
 group_id           | bigint                 |           |          | 
 patient_id         | bigint                 |           |          | 
 is_patient_related | integer                |           |          | 
 business_content   | json                   |           |          | 

his5_dms=# select count(1) from hiscrm.t_sys_oper_log;
 count 
-------
    97
(1 row)

查询表对应的具体文件

his5_dms=# SELECT oid,relfilenode FROM pg_class WHERE relname='t_sys_oper_log';
  oid  | relfilenode 
-------+-------------
 16850 |       16850
(1 row)

his5_dms=# SELECT pg_relation_filepath('hiscrm.t_sys_oper_log');
 pg_relation_filepath 
----------------------
 base/16386/16850
(1 row)

his5_dms=# SHOW data_directory;
     data_directory     
------------------------
 /var/lib/pgsql/12/data
(1 row)

使用dd对文件进行破坏

[postgres@xifenfeidg ~]$ ls -l  /var/lib/pgsql/12/data/base/16386/16850
-rw-------. 1 postgres postgres 90112 Sep  5 20:26 /var/lib/pgsql/12/data/base/16386/16850
[postgres@xifenfeidg ~]$ dd if=/dev/zero of=/var/lib/pgsql/12/data/base/16386/16850 bs=512 count=1 conv=notrunc
1+0 records in
1+0 records out
512 bytes copied, 0.000158756 s, 3.2 MB/s

重启pg库

[postgres@xifenfeidg bin]$ ./pg_ctl -m fast -D /var/lib/pgsql/12/data/ stop
waiting for server to shut down.... done
server stopped
[postgres@xifenfeidg bin]$ ./pg_ctl -D /var/lib/pgsql/12/data/ start
waiting for server to start....2025-03-02 19:02:11.395 HKT [64515] LOG:  
  starting PostgreSQL 12.20 on x86_64-pc-linux-gnu, 
  compiled by gcc (GCC) 8.5.0 20210514 (Red Hat 8.5.0-22), 64-bit
2025-03-02 19:02:11.396 HKT [64515] LOG:  listening on IPv6 address "::1", port 5432
2025-03-02 19:02:11.396 HKT [64515] LOG:  listening on IPv4 address "127.0.0.1", port 5432
2025-03-02 19:02:11.396 HKT [64515] LOG:  listening on Unix socket "/var/run/postgresql/.s.PGSQL.5432"
2025-03-02 19:02:11.397 HKT [64515] LOG:  listening on Unix socket "/tmp/.s.PGSQL.5432"
2025-03-02 19:02:11.403 HKT [64515] LOG:  redirecting log output to logging collector process
2025-03-02 19:02:11.403 HKT [64515] HINT:  Future log output will appear in directory "log".
 done
server started

查询数据报错

[postgres@xifenfeidg bin]$ psql
psql (16.8, server 12.20)
Type "help" for help.

postgres=# \c his5_dms;
psql (16.8, server 12.20)
You are now connected to database "his5_dms" as user "postgres".
his5_dms=#  select count(1) from hiscrm.t_sys_oper_log;
ERROR:  invalid page in block 0 of relation base/16386/16850

通过pdu进行恢复
跳过了坏块,把好的block中数据均恢复出来

his5_dms.hiscrm=# unload tab t_sys_oper_log;
正在解析表 <t_sys_oper_log>. 已解析数据页: 0, 已解析数据: 0 条
        |-块号0 空页面或页面已损坏,已跳过
正在解析表 <t_sys_oper_log>. 已解析数据页: 11, 已解析数据: 86 条
表名<t_sys_oper_log>-</var/lib/pgsql/12/data/base/16386/16850> 
    解析完成, 11 个数据页 ,共计 86 条数据. 成功 86 条; 失败【0】条 
COPY文件路径为:<his5_dms/hiscrm/t_sys_oper_log.csv>

导入pg库中

his5_dms=# truncate table hiscrm.t_sys_oper_log;
TRUNCATE TABLE
his5_dms=# \i his5_dms/COPY/hiscrm_copy.sql
SET
COPY 86
his5_dms=# 
发表在 PostgreSQL恢复 | 标签为 , , , , , | 留下评论

linux rm -rf 删除数据文件恢复

有客户由于误操作删除了oracle的部分数据文件(rm -rf 方式删除),然后自己尝试进行恢复操作,对部分文件执行了offline,导致比较麻烦的后果
jb
offline-file


接手故障之后,第一时间对其进行了镜像(因为有部分文件句柄已经释放,为了方式覆盖进一步破坏),对于没有释放的句柄可以通过类似方法进行恢复,参考以前类似恢复:
Oracle误删除数据文件恢复
Solaris rm datafile recovery—利用句柄误删除数据文件恢复

!cp  269  /u01/app/oracle/oradata/orcl/XXXXXX_DATA01.dbf
alter database datafile 12 offline;
recover datafile 12;
alter database datafile 12 online;

对于删除文件,而且句柄已经释放的文件,通过文件系统层面反删除进行恢复,参考以前类似恢复:
rm -rf误删Oracle数据库恢复
记录一次rm -rf 删除数据文件异常恢复
rm -rf 删除数据文件恢复方法—文件系统反删除+oracle碎片重组
rm


在这个恢复过程中,由于客户linux是物理机,而且本地空间不足,无法对其进行镜像,采用dd命令直接写镜像到其他的linux机器上(通过nfs方式),然后在win机器上直接挂载该nfs,记录下win上挂载nfs操作
nfs
mount-nfs

发表在 Oracle备份恢复 | 标签为 , , , | 留下评论