Point In Time Copy of IBM Spectrum Scale data
Most storage systems provide the functionality to make a point-in-time copy of data as an online backup mechanism. This function provides an instantaneous copy of the original data on the target disk, while the actual copy of data takes place asynchronously and is fully transparent to the user.
Several uses of the FlashCopy® replica after its initial creation can be considered. For example, if your primary operating environment suffers a permanent loss or a corruption of data, you may choose to flash the target disks back onto the originals to quickly restore access to a copy of the file system as seen at the time of the previous snapshot. Before restoring the file system from a FlashCopy, please make sure to suspend the activity of the GPFS™ client processes and unmount the file system on all GPFS nodes. FlashCopies also can be used to create a copy of data for disaster recovery testing and in this case are often taken from the secondary devices of a replication pair.
When a FlashCopy disk is first created, the subsystem establishes a control bitmap that is subsequently used to track the changes between the source and the target disks. When processing read I/O requests sent to the target disk, this bitmap is consulted to determine whether the request can be satisfied using the target's copy of the requested block. If the track containing the requested data has not yet been copied, the source disk is instead accessed and its copy of the data is used to satisfy the request.
To prevent the appearance of out-of-order updates, it is important to consider data consistency when using FlashCopy. When taking the FlashCopy image all disk volumes that make up the file system must be copied so that they reflect the same logical point in time. Two methods may be used to provide for data consistency in the FlashCopy image of your GPFS file system. Both techniques guarantee the consistency of the FlashCopy image by the means of temporary suspension of I/O, but either can be seen as the preferred method depending on your specific requirements and the nature of your GPFS client application.
- Import the state of the file system's configuration from the primary location.
- Propagate all relevant changes to the configuration in the primary
cluster to its peer to prevent the risks of discrepancy between the
peer's mmsdrfs file and the content
of the file system descriptor found in the snapshot.
It is suggested you generate a new FlashCopy replica immediately after every administrative change to the state of the file system. This eliminates the risk of a discrepancy between the GPFS configuration data contained in the mmsdrfs file and the on-disk content of the replica.