• Considerations for synchronous replication

    PDF

    On this page

    Considerations for synchronous replication

    Synchronization states

    Replication sessions configured with synchronous replication (RPO set to zero) have synchronization states that describe the state of the user data on the destination storage resource. These states are useful for troubleshooting. However, for replication session management, it is recommended that you check the operational status. For accurate information about the synchronization state of a replication, see the synchronization state on the source system.

    NOTE The synchronization state on the destination system is not updated after a synchronous replication session is fractured.

    Synchronous replication sessions can have the following states:

    • Consistent – The data on the destination system is consistent from the host perspective. The destination LUN is the latest or point in time copy of the source LUN.
    • In Sync – The data on the destination system is consistent from the host perspective. The destination LUN is an exact copy of the source LUN.
    • Syncing – The data on the destination system is updated from the source system. The updates represent the pending changes on the source system. These changes are not yet replicated to the destination system because of the following reasons:
      • A new replication session requires a full synchronization.
      • The replication session requires a full synchronization after an unplanned failover operation, when a failback or resume operation is performed.

        If there is no common base snapshot between the source system and the destination system, a full synchronization is required after an unplanned failover operation.

      • The replication session was previously in a Consistent or In Sync state. A connection issue has created a backlog of changes on the source system. The backlog is updated to the destination system once the connection issue is resolved.
      • The replication session was paused resulting in a backlog of changes on the source system. The backlog is updated to the destination system once the replication session is resumed.
    • Out of Sync – The replication session cannot determine the state of the data on the destination LUN. The update to the destination system may be incomplete. This issue can occur for any of the following reasons:
      • A connection issue disrupts a replication that is in a Syncing state.
      • The replication session is paused on the source system.
      • The replication session was deleted locally on the destination system.
      NOTE If a synchronous replication session is fractured or paused and there are still data writes on the source storage resource, you should recover or resume the session as soon as possible. If data writes are sparsely distributed on the source storage resource, resynchronizing may transport more data than the original data writes and, if there are snapshots on the destination storage resource, they occupy more storage space size. If those old snapshots are deleted, the size shrinks back.
    • Inconsistent – The state is reported only when replication is failed over. It appears when the synchronization state of the session was not In Sync or Consistent before failover. In this case, it is recommended that you check for the consistency of the destination storage resource.

    Synchronization state transitions

    The following table lists the synchronization state transitions for some of the replication actions:

    Action Synchronization state transition
    Initial synchronization Unknown > Out of Sync > Syncing > Consistent > In Sync
    Pause In Sync > Consistent or Syncing > Out of Sync
    Resume Consistent or Out of Sync > Syncing > Consistent > In Sync
    Fail over Syncing > Inconsistent or Out of Sync > Inconsistent
    NOTE If the state was In Sync or Consistent before a failover occurred, that state is retained.
    Replication connection down Syncing > Out of Sync or In Sync > Consistent (if there is I/O)
    NOTE For synchronous replication, I/O performance on a storage resource will be negatively impacted after a synchronous replication session is created on it. Also, initial synchronization might take a long time to complete depending on the storage object size and thin or thick type. For a thick storage resource, synchronous replication always copies the complete storage corresponding to its size even when it is empty.

    Code page translation of NFSv3 and FTP clients

    All Unity systems (OE version 4.2 and earlier) are configured by default to support NFSv3 and FTP clients configured for UTF-8. The code page on these Unity systems cannot be changed.

    With Unity OE version 4.3 and later, the code page of the Unity system can be changed through the svc_nas {<NAS_server_name> | all} -param -facility vdm -modify codepage -value <value> service command to UTF-8, 8859-1, or 8859-15. Code page 8859-15, which includes the most used western European characters, is an extension of the 8859-1 code page.

    NOTE The default code page in Unity systems does not need to be changed for either NFSv4 or SMB clients. NFSv4 clients support only UTF-8 and SMB clients support only Unicode.

    With Unity systems OE version 5.0.3 and later, the code page of the Unity system can be changed through the same service command to CP949 (EUC-KR Korean). The code page on earlier Unity systems (OE version 5.0.2 and earlier) cannot be changed to CP949. Refer to Unity Service Commands Technical Notes for more information.

    NOTE A failover from Unity systems OE version 5.0.3 with the code page set to CP949 to a system with an earlier OE version will revert the codepage to UTF-8. On failback, the code page must be reconfigured on the version 5.0.3 system.

    Fibre Channel configuration for synchronous replication

    Every replication session is tied to a Storage Processor (SP) via a replication interface. For synchronous replication, a Fibre Channel (FC) connection is used for data communication, and an IP interface used for management commands. Before creating a replication interface or connection, it is recommended that you configure the FC zoning.

    NOTE When using Single Mode SFP (to support a long range connection), SM cables along with an SM SFP connection on the switch are required as well as manually increasing the credits on the switch. In the case of direct connection between storage systems, both Synchronous Replication ports should use SM SFP along with SM cables. Refer to the Unity Hardware Information Guide for more information about SM SFP.

    Only one FC port can be configured and used for synchronous replication. Once an FC port is assigned, it will continue to act as the synchronous replication port and can never be changed.

    To determine the FC port used for synchronous replication, in the CLI console, run the command /remote/sys show -detail. Port information, similar to the following example, will appear in the output:

    Synchronous FC ports = spb_fc4, spa_fc4

    For more information, see the Unisphere CLI User Guide.

    NOTE RecoverPoint and the native Replication feature cannot simultaneously operate on the same volume or Fibre Channel port of the storage system. Even if you do not configure or use synchronous replication, do not configure the synchronous replication port for RecoverPoint. There is no such restriction on the Ethernet ports for iSCSI. You can configure RecoverPoint to use any available Ethernet port.

    Zoning configuration

    Ensure that you complete the following:

    • Configure the port as an F-port to avoid any potential issues, and remove other types of ports from the list of allowed port types.
    • For a synchronous replication connection, connect SPA port of the source system to the SPA port of the destination system. Similarly, connect SPB port of the source system to the SPB port of the destination system.
      NOTE Only a one to one FC connection is allowed for synchronous replication. One to more than one FC connection is not supported. Also, residual zones to an earlier source system used in synchronous replication should be removed.

      For example,

      • OB_H1450 SPA Port 4 <--> OB_H1443 SPA Port 4
      • OB_H1450 SPB Port 4 <--> OB_H1443 SPB Port 4
      NOTE If the FC connections are incorrectly crossed configured, for example:
      • OB_H1450 SPA Port 4 <--> OB_H1443 SPB Port 4
      • OB_H1450 SPB Port 4 <--> OB_H1443 SPA Port 4
      connect the FC connections correctly. After correcting the FC connections, reboot both SPA and SPB of the source system and the destination system.
    • The SPA and SPB ports on the same side (source or destination) must have the identical port configuration. However, the configuration of ports on the source system can be different from the port configuration on the destination system. For example, SPA Port 4 <--> SPA I/O Module 0 Port 0 and SPB Port 4 <--> SPB I/O Module 0 Port 0 are supported.
    • It is recommended to have a separate FC port for FC host access. But, the synchronous replication FC port can be shared with host access.
    • Although setting the port speeds to Auto works, it is recommended that you set the port speeds for the FC ports on both systems and the FC switch port.

    High availability consideration

    To ensure high availability for block synchronous replication, consider the following:

    • If the FC connection for an SP is disrupted, set the SP ownership for the source LUN to the other SP. SP ownership on the destination LUN automatically switches to the corresponding SP on the destination system.
    • If an SP on the destination system shuts down or is faulted (not in a normal state), the system will automatically change the SP ownership on the destination LUN. Ensure that you manually set the SP ownership for the source LUN to the other SP.
    • For file synchronous replication, if an SP failover, failback, or NAS server local move results in a synchronously replicated source NAS server and file systems' current SP that is different from the destination NAS server and file systems' current SP, the sessions will be fractured until both are back to the same source SP and destination SP. Unlike block synchronous replication, no automatic switch is triggered. You can perform a NAS server local move manually to make the current SPs of the source and destination match.
    Questions?
    We're here to help.
    From offering expert advice to solving complex problems, we've got you covered.