Perform full restore

This topic describes how to physically restore data by using an SQL statement or in OceanBase Cloud Platform (OCP). OceanBase Database supports tenant-level point-in-time full restore.

OceanBase Database supports restore in the same cluster or to a different cluster.

Perform full restore by using an SQL statement

After you execute the restore statement, the system restores the required macroblock data from the corresponding full and incremental backups at the backup destination to the destination tenant and then pulls and replays the transaction logs.

Prerequisites

The preparations for restore are completed. For more information, see Preparations for restore.

Procedure

1. Log on to the sys tenant as the root user.

2. Optional. Execute the following statement to set a restore password for the backup.

   You need to set a restore password for the backup only if a password is added during data backup.

   obclient> SET DECRYPTION IDENTIFIED BY 'password';
   

   In the statement, password must be replaced with the password added during data backup. If the password for full backup is different from the password for incremental backup, you must enter the password for full before the password for incremental backup and separate them with a comma.

   Example:

   obclient> SET DECRYPTION IDENTIFIED BY '******';
   
   
   obclient> SET DECRYPTION IDENTIFIED BY '******','******';
   

3. Execute the following statement to start a restore task:

   obclient> ALTER SYSTEM RESTORE <dest_tenant_name> FROM <source_tenant_name> at 'uri' UNTIL 'timestamp' WITH 'restore_option';
   

   The following table describes the parameters.

   Parameter	Description
   dest_tenant_name	The name of the destination tenant to which the data is to be restored.
   source_tenant_name	The name of the source tenant for which the data is backed up.
   uri	The value of backup_dest specified during backup.
   timestamp	The restore timestamp, which must be later than or equal to START_TIME in the CDB_OB_BACKUP_SET_DETAILS view of the earliest data backup, and earlier than or equal to MAX_NEXT_TIME in the CDB_OB_BACKUP_ARCHIVELOG_SUMMARY view of the log backup.
   restore_option	The restore options, including backup_cluster_name, backup_cluster_id, pool_list, locality, primary_zone, and kms_encrypt. backup_cluster_name: required. The name of the source cluster. backup_cluster_id: required. Same as the value of cluster_id of the source cluster. pool_list: required. The resource pool of the tenant. Multiple resource pools must be separated with commas (,). locality: optional. The locality of replicas of the new tenant, which must match the zone information of pool_list of the cluster where the new tenant is located. By default, the locality of the source tenant is used. If the value specified for this parameter does not match the zone information of pool_list of the cluster where the new tenant is located, the system resets the locality to the default value. That is, the system sets an F replica for each zone based on zone_list of resource_pool. For example, the locality of the source tenant is "F@z1,F@z2,F@z3", and the zone distribution of the cluster where the new tenant is located is "z1,z2,z3,z4", the locality of the source tenant is used as the locality of the new tenant. If the locality of the source tenant is "F@z1,F@z2,F@z3,F@z4,F@z5", and the zone distribution of the cluster where the new tenant is located is "z1,z2,z3", the locality of the new tenant covers all zones of this cluster. That is, the locality of the new tenant is "F@z1,F@z2,F@z3". primary_zone: optional. The preferred position of the leader of the new tenant, which must match pool_list and locality. That is, it must match the zone information and meet the constraint that primary_region must have at least two Paxos members. By default, the primary zone of the source tenant is used. If the value specified for this parameter does not match pool_list or locality of the cluster where the new tenant is located, the system resets primary_zone to the default value. That is, leaders are randomly distributed to the zones based on zone_list of locality. For example, if the locality of the source tenant is "F@z1,F@z2,F@z3", the primary zone is z1, and the zone distribution of the cluster where the new tenant is located is "z1,z2,z3,z4", the primary zone of the new tenant is also z1. If the locality of the source tenant is "F@z1,F@z2,F@z3,F@z4,F@z5", the primary zone is z1, and the zone distribution of the cluster where the new tenant is located is "z1,z2,z3", the locality of the new tenant is "F@z1,F@z2,F@z3" and the primary zone is "z1,z2,z3". kms_encrypt: optional. Specifies whether encryption information is required. If the value is true, the value of kms_encrypt_info specified in restore preparations is required for restore.

   Examples:

   • NFS

     obclient> ALTER SYSTEM RESTORE restored_trade FROM trade at 'file:///data/nfs/backup' until '2020-05-21 09:39:54.071670' with 'backup_cluster_name=xxx&backup_cluster_id=xxx&pool_list=xxx';
     

   • OSS

     obclient> ALTER SYSTEM RESTORE restored_trade FROM trade at 'oss://oceanbase-test-bucket/backup/?host=xxx.aliyun-inc.com&access_id=xxx&access_key=xxx' until '2020-03-23 08:59:45' with 'backup_cluster_name=ob20daily.backup&backup_cluster_id=1&pool_list=restore_pool';
     

   • COS

     obclient> ALTER SYSTEM RESTORE restored_trade from trade at 'cos://oceanbase-test-appid/backup?host=cos.ap-nanjing.myqcloud.com&access_id=xxx&access_key=xxx&appid=xxx' until ' 2020-03-23 08:59:45' with 'backup_cluster_name=ob20daily.backup&backup_cluster_id=1&pool_list=restore_pool';
     

   • OBS

     obclient> ALTER SYSTEM RESTORE restored_trade from trade at 'obs://oceanbase-test-appid/backup?host=obs.cn-east-2.myhuaweicloud.com&access_id=xxx&access_key=xxx&appid=xxx' until ' 2020-03-23 08:59:45' with 'backup_cluster_name=ob20daily.backup&backup_cluster_id=1&pool_list=restore_pool';
     

After you initiate a restore task, you can query the corresponding views to check the restore progress and results. For more information, see View the restore progress and results.

Perform full restore in OCP

You can also restore a cluster from a backup in OCP.

1. Log on to the OCP console.

2. Go to the Initiate Recovery page by using one of the following methods.

   • Method 1: Initiate a restore on the Recovery page.

     1. In the left-side navigation pane, choose Backup & Recovery > Recovery.

     2. On the Recovery page, click Initiate Recovery in the upper-right corner.

   • Method 2: Initiate a restore on the O&M page of a cluster.

     1. In the left-side navigation pane, click Clusters to go to the Clusters page.

     2. Click the target cluster in Clusters or Resource Usage to go to the Overview page of the cluster.

     3. In the left-side navigation pane, click Backup & Recovery.

     4. On the Backup & Recovery page of the cluster, click Initiate Recovery.

     Note

     You cannot restore the cluster by using Method 2 in the following circumstances:

     1. The cluster is not in the running state.

     2. The cluster does not have a backup strategy.

     OCP of this version does not support the restore of clusters without backup strategies by using Method 2.

3. Optional. Specify the storage configuration fields if you use Method 1.

   

   Parameter	Description
   Storage Type	You can select one of the following media: File: Network File System (NFS) media. OSS: Alibaba Cloud Object Storage Service (OSS) is supported only in OceanBase Database V2.2.76 and later. COS: Tencent Cloud Object Storage (COS) is supported only in OceanBase Database V2.2.77 BP2 and later.
   Storage Directory	Select the directory in which the backup files are stored. You can click Select Existing Configuration to automatically populate the following fields: Domain, User, AccessKey Pair, and Host for Parsing. You can also manually specify the fields.
   Domain	Enter the domain name of the cloud storage of backup files. If you select OSS in the Storage Type field, enter the domain name to access the file stored at OSS of Alibaba Cloud. If you select COS in the Storage Type field, enter the domain name to access the file stored at COS of Tencent Cloud, for example, cos.ap-shanghai.myqcloud.com.
   User	This parameter is displayed only when Storage Type is set to OSS. It is the user used to access the cloud storage of backup files.
   AccessKey Pair	This parameter is displayed only when Storage Type is set to OSS. It is the AccessKey pair used to access the cloud storage of backup files.
   Resource Identity (APPID)	When Storage Type is COS, move the pointer over the question mark (?) and specify the value as prompted.
   Project ID	When Storage Type is COS, move the pointer over the question mark (?) and specify the value as prompted.
   Project Key	When Storage Type is COS, move the pointer over the question mark (?) and specify the value as prompted.
   Host for Parsing	When Storage Directory is set to Select Existing Configuration, the host for parsing is automatically populated. If you manually enter a value for Storage Directory, the host for parsing is not automatically populated. You can manually select a host as prompted. Click Parse to obtain the list of clusters that can be restored.

4. Specify the parameters in the Recovery Source and Time section by selecting the values from the corresponding drop-down lists.

   Parameter	Description
   Source Cluster	If you initiate a restore task from the O&M page of a cluster, the cluster is the source cluster by default, you do not need to specify this parameter.
   Source Tenant	Select the tenant to restore.
   Recovery Date	Select a date within the specified restorable time range.
   Restoration Point in Time	Select a point in time within the specified time range.

   

5. Select the destination tenant of the restore task. Select a cluster in the Running state from the drop-down list and enter the new tenant name.

   The following types of clusters cannot be used as the restore destination:

   • Clusters in the primary/standby mode.

     You cannot select a cluster in the primary/standby mode because physically restored data cannot be synchronized to the corresponding standby or primary database. For a primary or standby database, only clog data can be synchronized.

   • Clusters in the Unavailable state.

6. Specify the following parameters in the Zone Configuration section and set the priorities of the zones.

   Parameter	Description
   Replica Type	Valid values: Full-featured Replica: A general-purpose replica that has a full set of data types and features including transaction logs, a MemTable, and an SSTable. A full-featured replica can switch to the leader role efficiently to provide services. Log Replica: A replica that contains only logs. It does not have a MemTable or an SSTable. It provides log services for external applications and participates in log voting. It can facilitate the restore of other replicas, but it cannot become the leader to provide database services. Read-only Replica: A replica that contains full logs, MemTables, and SSTables. You can select this type of replica for applications that does not require high consistency in data reading.
   Unit Specification	The specification of the resources that the replica occupies. You can select an existing specification or create a specification as needed and select the new one.
   Units	The amount of resource units occupied by the replica.

   

7. Click Initiate Recovery.