ApsaraDB for HBase allows you to back up and restore data at low costs and in an efficient manner.

Features and benefits

  • Synchronize incremental data in real time and achieve a recovery point objective (RPO) of less than 30 seconds.
  • By default, the entire database is backed up and single table recovery is supported.
  • Restore data at the scheduled time.

Scenarios

  • Back up and restore data in ApsaraDB for HBase databases.
  • Back up and restore data in user-created HBase databases running on Elastic Compute Service (ECS) instances.
  • Back up and restore data in on-premises databases and other HBase databases in the cloud.

Limits

  • Only ApsaraDB for HBase clusters can be restored. You can restore ApsaraDB for HBase V1.x, V2.x, and Performance-enhanced Edition.
  • You cannot back up or restore SOLR indexes.
  • Insufficient storage space interrupts data backup. However, unfinished backup files are retained.

Basic operations

  1. Enable data backup
    1. Log on to the ApsaraDB for HBase console.
    2. Click the name of the ApsaraDB for HBase cluster that you want to manage.
    3. Click Backup and Restoration.
    4. Click Enable Now.Enable
  2. In the left-side navigation pane, click Backup Configuration to create a backup plan.Backup Configrtion
    Note
    • Backup table: Support wildcard, if you want to back up the whole library, you can fill in it directly.
    • Full backup cycle(days): A full backup is triggered every 7 days by default. The optional time range is 3 to 10 days, too small may not be completed in a cycle, too long will affect the data recovery time.
    • Next full backup time: Time to start the next full backup.
    • Number of full backup reservations: Number of full backups reserved.
  3. Restore data. Click Create Recovery.

    Configure the following parameters:

    • Target Cluster

      The cluster that you want to restore.

    • Restore Time
    • Restore table
    Note Enter a table name each line. The table name cannot contain wildcard characters and must be in the format of namespace:table in which namespace is required. For example, default:testTable. You can also restore a table to another table by specifying namespace:table/namespace:table2. For example, default:testTable/default:testTable2. After the restoration task starts, you can view the progress on the Restoration Records page.

Estimate the storage space for backup and restoration

Storage space for backup and restoration = Storage space for full backup + Storage space for incremental backup

Storage space for full backup = (Number of full backup reservations + 1) × Size of a full backup reservation

The previous backup will not be deleted until the new backup is created. Therefore, extra full backup storage space is required during the backup.

You can check the storage space for full backup on the HBase Master page or the UE cluster management page.

Storage space for incremental backup = Number of log retention days × Size of daily incremental logs

Assume that data is written at an average speed of 1 MB/s. In this case, 84 GB of storage space is required each day. You can obtain the write speed of the cluster through the Monitoring and Alerts module, or contact the HBase DingTalk group. If a cluster receives a large amount of data every day, the size of daily incremental logs is large. You can reduce the interval for full backup to reduce the log size.

We recommend that you properly evaluate the storage usage. Insufficient storage space causes backup interruption.

What is the latest time point to which I can restore data?

If no pending tasks exist, the latest time point to which you can restore data depends on the interval for importing ApsaraDB for HBase log data to Object Storage Service (OSS). The default interval is every 30 seconds or every 30 MB. If a failure occurs, you may fail to restore the data written into ApsaraDB for HBase databases within 30 seconds before the current time.

How much time does the system require to restore data?

The time that the system requires to restore data depends on the data size and bandwidth.

  • Full data restoration: The maximum bandwidth of OSS is 1 GB per second and each BDS Worker writes 100 MB per second.
  • Incremental data restoration: Each node in a specified HBase cluster receives 30 MB to 40 MB per second and each BDS Worker writes 100 MB per second.

Can I restore a single table?

Yes, you can restore a single table. If duplicate tables exist in a database, you can restore the table to another table.