gp_dump备份恢复greenplum

       gp_dump是GP并行备份的备份工具，在运行gp_dump的时候master与所有的segment节点都开始备份，数据文件都是放在各种的节点服务器上，消耗的时间和数据量最大的、消耗时间最长的节点有关，这里就不全部介绍gp_dump所有的参数，可以自行gp_dump --help查看。下面只介绍下面实验用到的参数。

-h                     GPmaster主机名

-p                     GPmaster端口

-U                     用户名

-c                     备份包含数据定义删除语句

-d                     库名

--gp-d              备份的数据目录

--gp-r               备份report目录

测试：

一、创建测试数据

1、create table a as select id from generate_series(1,10000) id distributed by (id)

-- 创建10000条数据

2、创建几个function对象

二、备份

[root@gp-master ~]# gp_dump -h 172.30.248.242 -p 5432 -U gpadmin -d recover  --gp-d /disk/backup/ --gp-r /disk/backup/

.....

20161102:18:35:27|gp_dump-[FATAL]:-could not start Greenplum Database backup: ERROR:  could not create backup directory "/disk/backup/": Permission denied

20161102:18:35:27|gp_dump-[FATAL]:-could not start Greenplum Database backup: ERROR:  could not create backup directory "/disk/backup/": Permission denied

20161102:18:35:27|gp_dump-[FATAL]:-could not start Greenplum Database backup: ERROR:  could not create backup directory "/disk/backup/": Permission denied

20161102:18:35:27|gp_dump-[FATAL]:-could not start Greenplum Database backup: ERROR:  could not create backup directory "/disk/backup/": Permission denied

20161102:18:35:28|gp_dump-[FATAL]:-could not start Greenplum Database backup: ERROR:  could not create backup directory "/disk/backup/": Permission denied

20161102:18:35:28|gp_dump-[FATAL]:-could not start Greenplum Database backup: ERROR:  could not create backup directory "/disk/backup/": Permission denied

....

节点上没有目录权限

这个时候需要kill掉这个gp_dump进程

[root@gp-master backup]# ps -ef | grep gp_dump

gpadmin  27296 26783  0 18:35 pts/0    00:00:00 gp_dump -h 172.30.248.242 -p 5432 -d recover --gp-d /disk/backup/ --gp-r /disk/backup/

root     27386 27334  0 18:37 pts/1    00:00:00 grep gp_dump

[root@gp-master backup]# kill -9 27296

创建目录并给予权限

[root@gp-master ~]# gpssh -f /tmp/segment_key -e 'mkdir -p /disk/backup'

[gp-segment4] mkdir -p /disk/backup

[gp-segment5] mkdir -p /disk/backup

[gp-segment1] mkdir -p /disk/backup

[gp-segment2] mkdir -p /disk/backup

[gp-segment3] mkdir -p /disk/backup

[root@gp-master ~]# gpssh -f /tmp/segment_key -e 'chown -R gpadmin:gpadmin /disk/backup'

[gp-segment4] chown -R gpadmin:gpadmin /disk/backup

[gp-segment1] chown -R gpadmin:gpadmin /disk/backup

[gp-segment5] chown -R gpadmin:gpadmin /disk/backup

[gp-segment2] chown -R gpadmin:gpadmin /disk/backup

[gp-segment3] chown -R gpadmin:gpadmin /disk/backup

再次执行gp_dump备份

[root@gp-master ~]# gp_dump -h 172.30.248.242 -p 5432 -U gpadmin -d recover --gp-d /disk/backup/ --gp-r /disk/backup/

......

Greenplum Database Backup Report

Timestamp Key: 20161102184312

gp_dump Command Line: -h 172.30.248.242 -p 5432 -U gpadmin -d recover --gp-d /disk/backup/ --gp-r /disk/backup/

Pass through Command Line Options: -d

Compression Program: None

Backup Type: Full

Individual Results

        segment 7 (dbid 9) Host gp-segment4 Port 40001 Database recover BackupFile /disk/backup/gp_dump_0_9_20161102184609 : Succeeded

        segment 6 (dbid 8) Host gp-segment4 Port 40000 Database recover BackupFile /disk/backup/gp_dump_0_8_20161102184609 : Succeeded

        segment 5 (dbid 7) Host gp-segment3 Port 40001 Database recover BackupFile /disk/backup/gp_dump_0_7_20161102184609 : Succeeded

        segment 4 (dbid 6) Host gp-segment3 Port 40000 Database recover BackupFile /disk/backup/gp_dump_0_6_20161102184609 : Succeeded

        segment 3 (dbid 5) Host gp-segment2 Port 40001 Database recover BackupFile /disk/backup/gp_dump_0_5_20161102184609 : Succeeded

        segment 2 (dbid 4) Host gp-segment2 Port 40000 Database recover BackupFile /disk/backup/gp_dump_0_4_20161102184609 : Succeeded

        segment 1 (dbid 3) Host gp-segment1 Port 40001 Database recover BackupFile /disk/backup/gp_dump_0_3_20161102184609 : Succeeded

        segment 0 (dbid 2) Host gp-segment1 Port 40000 Database recover BackupFile /disk/backup/gp_dump_0_2_20161102184609 : Succeeded

        Master (dbid 1) Host gp-master Port 5432 Database recover BackupFile /disk/backup/gp_dump_1_1_20161102184609 : Succeeded

        Master (dbid 1) Host gp-master Port 5432 Database recover BackupFile /disk/backup/gp_dump_1_1_20161102184609 _post_data: Succeeded

gp_dump utility finished successfully.

到segment上看下备份文件

[root@gp-segment4 backup]# ls

gp_dump_0_8_20161102184609  gp_dump_0_9_20161102184609  gp_dump_status_0_8_20161102184609  gp_dump_status_0_9_20161102184609

注：segment上的备份文件格式是gp_dump_0_dbid_timestamp，master上的备份文件格式是gp_dump_1_dbid_timestamp，

这个timestamp在恢复的时候会用到

三、删除数据并恢复

删除数据

truncate table a;

恢复：

[root@gp-master ~]# gp_restore -h 172.30.248.242 -U gpadmin -d recover --gp-d /disk/backup/ --gp-r /disk/backup/ --gp-k=20161102184609

Greenplum Database Restore Report

Timestamp Key: 20161102184609

gp_restore Command Line: -h 172.30.248.242 -U gpadmin -d recover --gp-d /disk/backup/ --gp-r /disk/backup/ --gp-k=20161102184609

Pass through Command Line Options: None

Compression Program: None

Individual Results

        Restore of database "recover" on Master database: Failed with error:

{ERROR:  function "cdm_account" already exists with same argument types

20161102:19:17:41|gp_restore_agent-[ERROR]:-psql finished abnormally with return code 3.

20161102:19:17:41|gp_restore_agent-[ERROR]:-Finished with errors

}

报错，函数cdm_account已经存在，因为我们只删除了表数据，test库中其他对象没有删除，如果备份的时候带-c参数，备份数据时包含数据定义的删除语句，所以先恢复到别的库

[root@gp-master ~]# gp_restore -h 172.30.248.242 -U gpadmin -d test  --gp-d /disk/backup/ --gp-r /disk/backup/ --gp-k=20161102184609

...

gp_restore  utility finished successfully.

数据和函数对象都恢复成功

注：在恢复的时候，GP的segment必须与备份的时候一致

来自 “ ITPUB博客 ” ，链接：http://blog.itpub.net/29989552/viewspace-2127566/，如需转载，请注明出处，否则将追究法律责任。

转载于:http://blog.itpub.net/29989552/viewspace-2127566/