All Products
Search
Document Center

PolarDB:Restoration process and estimated time

Last Updated:Mar 22, 2024

You can restore an entire cluster by using either of the following two methods: restore data to a previous point in time or restore data from a backup set (snapshot).

The difference between the two methods is whether or not the point in time that you want to restore data to is the time that a backup set was created. If the point in time that you want to restore data to is when a backup set was created, you can choose to restore data from a backup set (snapshot). If no backup sets were created at the point in time that you want to restore data to, you can only choose to restore data to a previous point in time.

Restoration process

The restoration process is the same for all methods.

  1. First, you need to restore all of the data to a new cluster.

  2. Second, you need to verify the accuracy of the data in the new cluster.

  3. Then, you can migrate the restored data to the original cluster or switch your business to the new cluster.

Estimated time required for restoration

The time required for data restoration depends on the specifications and size of your cluster and the size of the backup set. The following table shows the estimated time required for data restoration.

Procedure

Estimated time

Restore data to a new cluster

Create a temporary cluster

5 to 10 minutes

Restore data from a backup set (snapshot)

3 GB/minute

Restore data by using redo logs

Note

Redo logs are used only when you want to restore data to a previous point in time.

1.5 GB/minute

Compare data in the new cluster

3 to 5 minutes

Migrate the restored data from the new cluster to the original cluster

Cluster size (in KB) × Maximum IOPS × 15% × 16 KB/second

For example, if the sizes of a cluster (4 cores, 16 GB memory, maximum IOPS of 32,000), backup set (snapshot), and redo logs are 100 GB, 10 GB, and 3 GB, you can use the following formulas to calculate the required restoration time:

  • Time required to restore the data of the backup set (snapshot): 10/3 = 3.3 minutes

  • Time required to restore data by using redo logs: 3/1.5 = 2 minutes

  • Time required to migrate the restored data from the new cluster to the original cluster: 100 × 1,024 × 1,024/(32,000 × 15% × 16) = 22.7 minutes

If you choose to restore data to a new cluster and then migrate the restored data from the new cluster to the original cluster, the overall required time is 5 + 3.3 + 2 + 3 + 22.7 = 36 minutes.

If you choose to restore data to a new cluster and then switch your business to the new cluster, the overall required time is 5 + 3.3 + 2 + 3 + time required for business switching + new cluster configuration time = 13 minutes + time required for business switching + new cluster configuration time.

Note

We recommend that you restore data during off-peak hours.