Help Center/ Object Storage Migration Service/ Best Practices/ Migration from Other Clouds to Huawei Cloud/ Migrating Data from Amazon S3 to Huawei Cloud OBS/ Background
Updated on 2024-12-19 GMT+08:00
View PDF
Share

Background

This tutorial walks you through the process of migrating data from Amazon Simple Storage Service (S3) to Huawei Cloud OBS.

As an online data migration service, Huawei Cloud OMS helps you migrate data from third-party object storage systems to Huawei Cloud OBS or between OBS buckets.

When you use OMS, you only need to specify a source data address and a destination OBS data address on the console, and then create a migration task or migration task group. For details about the differences between a migration task and a migration task group, see What Are the Application Scenarios of Migration Tasks and Migration Task Groups? After the migration task starts, you can view and manage it on the console.

Service Cutover

If service cutover is involved, see Service Cutover for more information.

You need to choose a cutover method based your requirements. The preceding solutions are for reference only.

Constraints

Use of OMS is limited as follows.

Table 1 OMS constraints

Item

Constraints

Region

OMS does not support cross-border data migration. You can only migrate data from AWS China regions to Huawei Cloud Chinese Mainland regions or from AWS Global regions to Huawei Cloud International regions. You need to check where your data source is and specify the correct account.

NOTE:

If you use an AWS Global account to migrate data in an AWS China region or use an AWS China account to migrate data in an AWS Global region, an error is reported when the source buckets are listed.

Source access mode

OMS supports AWS S3 domain-style access (for example, <your-bucket>.s3.amazonaws.com) and does not support path-style access (for example, http://s3.amazonaws.com/<your-bucket>).

Objects with multiple versions

By default, OMS migrates only the latest version of objects in source buckets.

Storage class of destination buckets

The storage class of destination buckets can only be standard or infrequent access. You can change the storage class of destination buckets after the migration is complete.

Migration network

Only migrations over the Internet are supported.

Metadata migration

Only Chinese characters, English characters, digits, and hyphens (-) can be migrated. Other characters cannot be migrated.

  • Chinese characters are URL encoded during the migration.
    CAUTION:

    Chinese punctuation marks cannot be URL encoded during the migration. If metadata contains Chinese punctuation marks, the corresponding object will fail to be migrated.

  • English characters, digits, and hyphens (-) are directly migrated without code conversion.

Migration scope

A single migration task or migration task group can only migrate data of one bucket. If data of multiple buckets needs to be migrated, you need to create multiple tasks or task groups.

Migration speed

Generally, OMS can migrate 10 TB to 20 TB of data per day. If you need a higher migration speed, you are advised to use storage migration workflows on MgC. MgC allows you to migrate data using dedicated, scalable migration clusters and up to 20 Gbit/s of bandwidth.

However, the speed depends on the number and size of source objects and the transmission distance over the Internet between the source and destination buckets. You are advised to create a migration task to test the migration speed. The maximum migration speed is five times the average speed of a single task because up to five tasks can be executed concurrently in a region by default. If you need to define a higher number of concurrent tasks, you can create a storage migration workflow on MgC.

Archived data

You need to restore archived data before the migration. Note that when there is archived data to be migrated, you need to:

  • Create migration tasks after the restoration is complete.
  • Configure a validity period for restored data based on the total amount of data to be migrated. This helps prevent migration failures because restored data becomes archived again during the migration.
  • Pay your source cloud vendor for restoring archived data. To learn about the pricing details, contact your source cloud vendor.

Migration tasks

Up to five concurrent migration tasks per account per region

NOTE:

If your destination regions are CN North-Beijing1 and CN South-Guangzhou, you can run up to 10 migration tasks concurrently.

Up to 5,000 migration tasks within 24 hours per account per region

Migration task groups

Up to five concurrent migration task groups per account per region

NOTE:

If your destination regions are CN North-Beijing1 and CN South-Guangzhou, you can run up to 10 migration tasks concurrently.

Synchronization tasks

Synchronization tasks share quotas with migration tasks and migration task groups, but enjoy a higher priority.

Up to five concurrent synchronization tasks per account per region

Object list file
  • An object list file cannot exceed 1,024 MB.
  • An object list file must be a .txt file, and the ContentType metadata of the file must be text/plain.
  • An object list file must be in UTF-8 without BOM.
  • Each line in an object list file can contain only one object name, and the object name must be URL encoded.
  • Spaces are not allowed in each line in an object list file. Spaces may cause migration failures because they may be mistakenly identified as object names.
  • Each line in an object list file cannot be longer than 65,535 characters, or the migration will fail.
  • The ContentEncoding metadata of an object list file must be left empty, or the migration will fail.

URL list file
  • A URL list file cannot exceed 1,024 MB.
  • A URL list file must be a .txt file, and its metadata Content-Type must be text/plain.
  • A URL list file must be in UTF-8 without BOM.
  • Each line in a URL list file can contain only one URL and one destination object name.
  • Each line in a URL list file cannot be longer than 65,535 characters, or the migration will fail.
  • The ContentEncoding metadata of a URL list file must be left empty, or the migration will fail.
  • Spaces are not allowed in each line in a URL list file. Spaces may cause migration failures because they may be mistakenly identified as object names.
  • In a URL list file, you must use a tab character (\t) to separate the URL and destination object name in each line. The format is [URL][Tab character][Destination object name]. Chinese and special characters in the source and destination object names must be URL encoded. For example: 
     http://xxx.xxx.xxx.xxx.com/doc/thefile2.txt doc/thefile2.txt http://xxx.xxx.xxx.xxx.com/the%20file.txt the%20file.txt 
http://xxx.xxx.xxx.xxx.com/the%20file2.txt the+file2.txt 
http://xxx.xxx.xxx.xxx.com/doc/thefile.txt doc/thefile.txt
    CAUTION:

    The URL encoding must start from the second character after the domain name in a line. Do not encode the protocol header, domain name, or slash before or after the domain name. Otherwise, the format verification will fail.

    In each line, use a tab character (Tab key on the keyboard) to separate the URL and the destination object name. Do not use spaces.

    In the preceding examples, after the files represented by the URLs are copied to the destination bucket, they are named doc/thefile2.txt, the file.txt, the file2.txt, and doc/thefile.txt.

  • URLs in a list file can be accessed using HEAD and GET methods.

Failed object list file

A maximum of 100,000 failed objects can be recorded in a failed object list file.

NOTE:

If more than 100,000 objects fail to be migrated in the migration task, you are advised to rectify the fault based on the existing failed object list and perform the migration again.