No relevant resource is found in the selected language.

This site uses cookies. By continuing to browse the site you are agreeing to our use of cookies. Read our privacy policy>Search

Reminder

To have a better experience, please upgrade your IE browser.

upgrade

HyperMirror Feature Guide

OceanStor V3 Series V300R006

This document is applicable to OceanStor 2200 V3, 2600 V3, 5300 V3, 5500 V3, 5600 V3, 5800 V3, 6800 V3, 18500 V3, and 18800 V3. This document describes the implementation principles and application scenarios of the HyperMirror feature. Also, it explains how to configure and manage HyperMirror.
Rate and give feedback:
Huawei uses machine translation combined with human proofreading to translate this document to different languages in order to help you better understand the content of this document. Note: Even the most advanced machine translation cannot match the quality of professional translators. Huawei shall not bear any responsibility for translation accuracy and it is recommended that you refer to the English document (a link for which has been provided).
Working Principle

Working Principle

This section describes the concepts and working principles behind HyperMirror.

Concepts

  • Data organization

    OceanStor storage system uses a storage virtualization technology, and the virtualized storage data comprises two parts: metadata volumes and data volumes.

    • A metadata volume records the data storage location.
    • A data volume saves users' data.
  • Local LUN

    A LUN in a local storage system.

  • External LUN

    A LUN in a heterogeneous storage system.

  • Mirror LUN

    A LUN that is converted from a local or external LUN and inherits its data and services. When either mirror copy of a mirror LUN becomes unavailable, services running on the mirror LUN are not affected.

  • Mirror copy

    A copy of a mirror LUN. Mirror copies are LUNs that cannot be mapped to hosts. After mirror copies are removed, storage space will be released to a storage pool.

  • Dual write

    Write I/Os delivered from a host to a mirror LUN are written to both mirror copies. After data is successfully written to both mirror copies, a response is returned to the host, indicating that the write is complete.

  • Round-robin read

    If the mirror copies of a mirror LUN contain the same data, a host can read data from the two mirror copies in an alternate manner.

  • Data Change Log (DCL)

    A DCL records data changes made to a mirror copy. A DCL is stored in the metadata space of a mirror copy.

NOTE:

The storage space for DCL is provided by all the disks in the disk domain where the mirror copy locates, and four-disk RAID 1 is used for data protection.

Working Principle

HyperMirror creates a mirror LUN for a local or external LUN and provides two mirror copies with the same data for the mirror LUN. If a mirror copy fails, a host can still access the mirror LUN and services on the host are not affected. After the mirror copy recovers, it will automatically synchronize data from the other mirror copy, enabling data consistency between the two mirror copies and providing data protection for the mirror LUN.

The HyperMirror implementation process includes three operations: HyperMirror task creation, splitting, and synchronization.

Creating a HyperMirror Task

Figure 1-1 illustrates the process of creating a HyperMirror task.

Figure 1-1 HyperMirror implementation process

  1. Before a mirror LUN is created:

    A host reads data from the LUN whose ID is 0 and writes data into it.

  2. Creating a mirror LUN:

    After you create a mirror LUN for the LUN, the storage system automatically generates a mirror LUN whose ID is 1 and the LUN becomes mirror copy A. The data volume of the mirror LUN is provided by the mirror copy. Therefore, no new data space is occupied.

    The LUN information exchange technology is used to exchange LUN ID information between the mirror LUN and mirror copy A. In this way, the host can use the mirror LUN to read data from the mirror copy and write data into it. The read and write processes of service data are transparent to the host. The LUN ID of the mirror LUN becomes 0 and the LUN ID of the mirror copy becomes 1, as shown in Figure 1-1. The host continues reading data from the LUN whose ID is 0 and writing data into it, that is, the host reads data from the mirror LUN and writes data into it.

  3. Adding a mirror copy:

    Select a different disk domain and add a mirror copy (mirror copy B) to the mirror LUN. Mirror copy B is a LUN that is automatically generated by the storage system. A synchronization relationship is established between mirror copy B and mirror copy A. Data of mirror copy A is fully synchronized to mirror copy B. A HyperMirror task is created successfully.

Data Read/Write Process

  • Data read process

    When the host sends a read request to the mirror LUN after the HyperMirror task is created, the mirror LUN sends read I/Os to the two mirror copies in round-robin mode, that is, round-robin read is executed on the two mirror copies, as shown in Figure 1-2. After data is read from one mirror copy successfully, the read result is returned to the mirror LUN. Then the mirror LUN returns the read result to the host.

Figure 1-2 Data read process

You can set a read priority for two mirror copies respectively. If the read priority of one mirror copy is the same as that of the other mirror copy, the host reads data from the mirror copies in round-robin mode. If the read priority of one mirror copy is higher, the host directly reads data from the mirror copy. If the mirror copy with the higher read priority fails and data cannot be read from it, the host reads data from the other mirror copy, ensuring non-stop host services.

  • Data write process

    When the host sends a write request to the mirror LUN after the HyperMirror task is created, the mirror LUN sends write I/Os to the two mirror copies in dual-write mode, that is, dual-write operation is executed on the two mirror copies, as shown in Figure 1-3. After data is written into two mirror copies successfully, the write result is returned to the mirror LUN. Then the mirror LUN returns the write result to the host. Data is written into both mirror copies, ensuring data consistency between the mirror copies.

Figure 1-3 Data write process

If data fails to be written into one mirror copy, the status of the mirror copy becomes interrupted unexpectedly. The read and write I/Os of the host will not be sent to the mirror copy but will be sent to the other mirror copy. Therefore, host services are not affected.

Splitting

Splitting involves the following two scenarios:

  • If a mirror copy is an external LUN and you want to upgrade its storage system, you must split the external LUN before upgrading its storage system.
  • If the poor performance of one mirror copy adversely affects the overall LUN performance, you must split the LUN.

Mirror copies are split at a point in time and data synchronization between them is interrupted, as shown in Figure 1-4. After the mirror copies are split, I/Os of the mirror LUN are not sent to mirror copy A that is split but to mirror copy B. Therefore, host services are not affected. Because data synchronization between mirror copy A and mirror copy B is interrupted, data changes of the two mirror copies will be recorded in a DCL for the data synchronization next time.

Figure 1-4 Splitting process

Synchronization

Synchronization involves initial synchronization and incremental synchronization, as shown in Figure 1-5.

Figure 1-5 Synchronization process

  • Initial synchronization

    After a HyperMirror task is created, data of mirror copy A is fully copied to mirror copy B, ensuring data consistency between the two mirror copies.

    If no data is read from a LUN or written into the LUN before a HyperMirror task is created, you do not need to perform initial synchronization.

  • Incremental synchronization

    When mirror copy A recovers from Split or Interrupted to Normal, the storage system will copy incremental data from mirror copy B to mirror copy A. After incremental synchronization is complete, the storage system will automatically clear the data changes from the DCL.

Translation
Download
Updated: 2019-07-12

Document ID: EDOC1000138859

Views: 14520

Downloads: 247

Average rating:
This Document Applies to these Products
Related Documents
Related Version
Share
Previous Next