Running Global Mirror for an unplanned failover and failback

Edit online

Use this process to manage an unplanned failover and failback for the local (production) and remote (recovery) sites.

About this task

Global Mirror provides two-site extended distance remote copy disaster recovery. When a disaster occurs at the local site, you must initiate the failover and failback recovery of consistent data on the remote site. Host activity can resume on the local site when the host recovers but not before a consistent set of data is copied to all primary volumes on the local site.

With Global Mirror, the data that the host writes to the storage system at the local site is asynchronously shadowed to the storage system at the remote site. A consistent copy of the data is then automatically maintained on the storage system at the remote site.

The use of Global Mirror does not prevent all cases of data loss. During a disaster, data can be restored only to the last known consistent increment that was created. Data that is written to the primary site and awaiting transfer to the secondary site is lost if the two storage systems can no longer communicate.

The following considerations apply when you use the Global Mirror recovery process:
  • The Global Mirror master might still be running at the local site, especially if the disaster at the local site is a rolling disaster in which not all components fail simultaneously.
  • The consistent copy at the remote site is not the secondary volume, but it is the FlashCopy target whose source is the secondary volume.
  • The in-progress formation of a consistency group can stop at the time of the failure.
  • You can speed up recovery processing by choosing the Fast Reverse restore process that is explained later.

Procedure

Complete the following steps to use Global Mirror for an unplanned failover.

  1. Enter the Global Mirror session at the local site.
    Note: Before the next step, await until the master storage system completes processing or enters the unrecoverable state. If the local site failed, proceed to the next step without waiting.
  2. Initiate a recovery failover on the Global Copy volumes pair to force stoppage of the volume A to B extended distance relationship. Then, create a volume B to A Global Copy relationship.
    Note: All B volumes must successfully process the recovery failover request before you can move to the next step.
  3. Look at the session properties for volumes B and C to ascertain the state of the consistency group between the B and C volumes. You are looking primarily at the FlashCopy relationships and your analysis determines your next step in the recovery process. Act on your analysis as follows:
    1. FlashCopy relationships are not restorable and all the sequence numbers are equal. No action to the consistency group is necessary.
    2. FlashCopy relationships are restorable and all the sequence numbers are equal. Initiate the FlashCopy Discard changes action for all the FlashCopy relationships in the consistency group.
    3. All the FlashCopy sequence numbers are equal and at least one of the FlashCopy relationships is not restorable. Initiate the FlashCopy Commit changes action for all the FlashCopy relationships in the consistency group that are restorable.
    4. You have a mixed list of FlashCopy relationships; some are restorable and some are not restorable. The sequence numbers of restorable relationships have the same sequence number. Relationships that are not restorable have equal sequence numbers that differ from restorable relationships. Initiate the FlashCopy Commit changes action for all the FlashCopy relationships in the consistency group that are restorable.
    5. You have a mixed list of FlashCopy relationships; some are restorable and some are not restorable. The sequence numbers are not the same within each type of relationship. In this case, the recovery plan cannot continue because the Global Mirror process is corrupted. If the Global Mirror process is corrupted, you must recover your data through your last good backup.
    Note: When the state of all the FlashCopy relationships are known, consider initiating a tape backup of Volume C.
  4. Initiate the fast reverse restore process from the C volumes to the B volumes, selecting the Initiate background copy option.
    Notes:
    1. When you initiate the fast reverse restore process, Volume C becomes unusable.
    2. I/O must not be allowed to the B or C volumes during the fast reverse restore process.
    3. If you do not want to use the fast reverse restore process, use the Recovering from a disaster without using fast reverse restore procedure instead of this step.
  5. Before the next step, await completion of the background copy. The C to B FlashCopy relationship ends when the background copy completes.
  6. Initiate FlashCopy from volume B back to C. Ensure that you also select the Enable Change Recording and Inhibit writes to target volume options. This action creates a backup copy of the consistency group before applications begin to update the B volumes.
  7. Start the host I/O at the remote site on the B volumes. Production can continue on the remote site in this configuration until you are ready to return production to the local site.
  8. When you are ready to return production to the local site, run the recovery failback (B to A) to resynchronize the A volumes. The application at the remote site remains active.
  9. After the resynchronize process completes its first pass, quiesce the applications at your remote site so that the resynchronization can complete.
  10. When the resynchronization completes (no out-of-sync tracks), run recovery failover and failback with Global Copy on Volume A to re-create the Volume A to Volume B Metro Mirror relationship.
  11. Start the host I/O at the local site on the A volumes.
  12. Resume the Global Mirror process.