划线反馈

Restore 单表失败

更新时间：2026-05-13 07:31

适用版本：V2.1.x、V2.2.x、V3.1.x、V3.2.x内容类型：Troubleshoot

适用版本

OceanBase 数据库 V2.x、V3.X 版本。

问题描述

使用黑屏单表恢复租户。

create resource unit unit_2c8g max_cpu=2, min_cpu=2, max_memory='8g', min_memory='8g', max_iops=10000, min_iops=1000, max_session_num=1000000, max_disk_size='500g'; 

create resource pool restore_pool_rst unit = 'unit_2c8g', unit_num = 1; 

alter system set restore_concurrency = 50; 

alter system restore db1.tb1 for rst_oracle from ob_oracle at 'cos://' until '2023-07-21 22:40:18' with 'backup_cluster_name=obcluster&backup_cluster_id=1637xxxxxx&pool_list=restore_pool_rst&locality=F@zone1,F@zone2,F@zone3&primary_zone=zone1,zone2,zone3';

恢复失败，查看 restore history 表，显示 RESTORE_FAIL。

obclient [oceanbase]> select * from oceanbase.CDB_OB_RESTORE_HISTORY where JOB_ID=483\G;
*************************** 1. row ***************************
                   JOB_ID: 483
          EXTERNAL_JOB_ID: -1
                TENANT_ID: 1184
              TENANT_NAME: rst_oracle
         BACKUP_TENANT_ID: 1001
       BACKUP_TENANT_NAME: ob_oracle
        BACKUP_CLUSTER_ID: 1637xxxxxx
      BACKUP_CLUSTER_NAME: obcluster
               WHITE_LIST: `db1`.`tb1`
                   STATUS: RESTORE_FAIL
               START_TIME: 2023-07-27 18:15:50.268125
          COMPLETION_TIME: 2023-07-27 18:20:19.914677
          PARTITION_COUNT: 128
        MACRO_BLOCK_COUNT: -1
   FINISH_PARTITION_COUNT: 128
 FINISH_MACRO_BLOCK_COUNT: -1
  RESTORE_START_TIMESTAMP: 2023-07-19 05:00:05.272017
 RESTORE_FINISH_TIMESTAMP: 2023-07-21 22:40:18.000000
RESTORE_CURRENT_TIMESTAMP: NULL
          BACKUP_SET_LIST: cos://***ata/xxxxxxxx
        BACKUP_PIECE_LIST: cos://***csp.xxxxxxxx&appid=xxxxxx?host=c
                     INFO: ROOTSERVICE : OB_NOT_SUPPORTED(-4007) on "xxx.xxx.xxx.xxx:2882" with traceid YBxxxxxxxxxx-0005F7xxxxxxxxxx
1 row in set (0.005 sec)

可以看到，报错 database not exist，db1 这个库不存在。

[2023-07-27 18:20:19.903427] WARN  [RS.RESTORE] gen_white_list (ob_restore_scheduler.cpp:1886) [299954][752][YBxxxxxxxxxx-0005Fxxxxxxxxxx] [lt=12] [dc=0] database not exist(ret=-4007, ret="OB_NOT_SUPPORTED", item={database_name:"db1", table_name:"tb1"})

对比源库信息，源库库名为大写 DB1。

问题原因

数据库名和表名在黑屏恢复时需要保持大小写一致。否则会报 database not exist。

解决方案

修改黑屏执行命令。

alter system restore DB1.TB1 for rst_oracle from ob_oracle at 'cos://' until '2023-07-21 22:40:18' with 'backup_cluster_name=obcluster&backup_cluster_id=1637568272&pool_list=restore_pool_rst&locality=F@zone1,F@zone2,F@zone3&primary_zone=zone1,zone2,zone3';

恢复成功 RESTORE_SUCCESS。

obclient [oceanbase]> select * from oceanbase.CDB_OB_RESTORE_HISTORY where JOB_ID=485\G
*************************** 1. row ***************************
                   JOB_ID: 485
          EXTERNAL_JOB_ID: -1
                TENANT_ID: 1185
              TENANT_NAME: rst_oracle
         BACKUP_TENANT_ID: 1001
       BACKUP_TENANT_NAME: ob_oracle
        BACKUP_CLUSTER_ID: 1637xxxxxx
      BACKUP_CLUSTER_NAME: obcluster
               WHITE_LIST: `DB1`.`TB1`
                   STATUS: RESTORE_SUCCESS
               START_TIME: 2023-07-27 20:06:24.428165
          COMPLETION_TIME: 2023-07-27 20:17:21.400813
          PARTITION_COUNT: 1
        MACRO_BLOCK_COUNT: -1
   FINISH_PARTITION_COUNT: 1
 FINISH_MACRO_BLOCK_COUNT: -1
  RESTORE_START_TIMESTAMP: 2023-07-19 05:00:05.272017
 RESTORE_FINISH_TIMESTAMP: 2023-07-21 22:40:18.000000
RESTORE_CURRENT_TIMESTAMP: NULL
          BACKUP_SET_LIST: 
        BACKUP_PIECE_LIST: 
                     INFO: 
1 row in set (0.002 sec)

快速恢复的租户限制

Transfer 导致物理恢复完成后某些副本不可读

OceanBase 数据库