You can use a cloud backup solution such as Migrate to Virtual Machines to migrate the system to the Google Cloud environment by migrating the hosts and volumes (that are connected to the hosts) from the on-premise environment to the Compute Engine and Google Cloud Hyperdisk in the Google Cloud environment.
Data migration of the Cloud model for Google Cloud for VSP One SDS Block is an operation that migrates data from external storage to Google Cloud Hyperdisk, and then migrates the user data on that Google Cloud Hyperdisk to VSP One SDS Block in units of volumes. For Migrate to Virtual Machines, follow the Google Cloud operation guide.
Even if a failure occurs in the VSP One SDS Block storage system during data migration, the user data is secured.
The outline of the data migration procedure is as follows:
-
Add storage nodes for data migration (hereinafter, called initiator node) to VSP One SDS Block.
-
Connect the Google Cloud Hyperdisk to the initiator node.
-
Issue a REST API for data migration.
-
Remove the initiator node.
This document explains the data migration procedure using the following terminology.
- Google Cloud Hyperdisk
- Google Cloud service for providing storage volumes to be used in the Compute Engine instance.
- Compute Engine
- Google Cloud service to be used for starting and managing instances of Linux or Windows servers.
- External storage
- Storage system that stores the user data to be migrated.
- External volume
- Information about operation targets created on VSP One SDS Block for users to perform operations for the Google Cloud Hyperdisk (to which data is migrated from an external storage) by using VSP One SDS Block.
- Initiator node
- Storage node which is temporarily added to VSP One SDS Block for data migration.
- Migration destination server
- Newly installed server to use the user data migrated to VSP One SDS Block. Equivalent to a compute node in VSP One SDS Block.
- Migration destination volume
- Copy of the Google Cloud Hyperdisk that is created in the storage node by data migration. volumeType (volume type) of the volume is "MigrationDestination." When data migration completes, volumeType becomes "Normal."
- Migration source server
- Node in which the user application runs for I/O operations of user data in the external storage.
- Virtual volume
- Copy of the Google Cloud Hyperdisk that is temporarily created in the initiator node by data migration. volumeType of the volume is "ExternalMigrationOrigin."
Prerequisites of data migration
The prerequisites of data migration are as follows:
-
You need to migrate data to the Google Cloud Hyperdisk by using a cloud backup solution beforehand.
If the storage cluster applies Multi-Zone configuration, Zones of data migration destination must be the same for all Google Cloud Hyperdisks. *
Specify the Zone in which the storage cluster manages user data when migrating data from the Google Cloud Hyperdisk to the storage cluster.
* When selecting Zones, verify that the Zone where data is migrated (by using a cloud backup solution) and the Zone in which the storage cluster manages user data are the same. If there are multiple Zones for managing user data, select the Zone that has the largest total user data amount to be migrated.
-
As the initiator node, a server in which only the system drive is installed is supported. A server in which both the system drive and a user data drive are installed is not supported.
-
Data migration to a virtual private storage (VPS) in a multi-tenancy configuration is not supported.
-
QoS setting for a volume under data migration is not supported.
Operations that cannot be performed in parallel
The following cannot be performed during data migration:
-
Software update operations
-
Removal of migration-destination storage nodes
-
Maintenance blockage of the initiator node in which an external volume (whose status is "Normal") exists
-
Change of virtual volume settings (editing the settings, expanding the virtual volume, creating a snapshot, and so on.)
Related operations
The operations related to data migration are as follows:
|
Category |
Reference |
|---|---|
|
Performing data migration |
|
|
Obtaining the progress of data copy and external volume information |
|
|
Stopping data migration |
Notes on data migration
-
When to perform data migration:
Perform data migration when the I/O load is low (for example, nights and holidays) due to the following reason.
Once virtual volumes are created in the initiator node, I/O operations on virtual volumes can be performed from the migration destination server. However, I/O response from the migration destination server to the virtual volumes is slow during data copy.
Also, because the internode network of VSP One SDS Block is heavily used during data copy, performance of ordinary volumes might be affected.
-
In case that I/O is performed from the migration destination server to the virtual volume during data copy:
-
If any of the following failures occurs during data copy, data copy and I/O to the virtual volume stop. To resume the I/O to the virtual volume, you need to handle the failure, and then reperform data copy.
Failure in the initiator node
Failure in the storage node
Failure in the user data drive
-
Maintenance blockage, maintenance recovery, or replacement of the storage node can be performed in parallel with data copy. However, if you perform maintenance recovery and replacement of the storage node during data copy, data copy and I/O to the virtual volume stop. Perform maintenance blockage, maintenance recovery, or replacement of the storage node before starting data copy or after finishing data copy as much as possible.
- If you perform I/O to the virtual volume during data copy, you cannot stop data migration to change operations back to the migration source server. If you perform I/O to the virtual volume during data copy, perform all steps of data migration until the end.
-
-
In case that data migration is performed for a large volume:
Once data copy is stopped, data copy must be performed again from the beginning for the volumes for which data copy is not complete. Especially note when you perform data migration for a large volume because data copy takes time for a large volume. For details about the estimated processing time of data copy, see Estimated processing time.