PRIMECLUSTER Global Disk Services Configuration and Administration Guide 4.1 (Linux)

6.3 Backing Up and Restoring through an External Server

This section discusses the method of backing up data from and restoring data back to logical volumes (called application volumes in this manual) in a shared class in the primary domain through a server in a domain different from the primary domain (called an external backup server in this manual).

The backup and restore operations through an external backup server can be categorized into 3 patterns.

1. Backing up and restoring a logical volume with no replication

   For details, see "6.3.1"

2. Backing up and restoring through snapshot by slice detachment

   For details, see "6.3.2"

3. Backing up and restoring using snapshots from a proxy volume

   For details, see "6.3.3"

The following table summarizes characteristics of the respective operations.

Patterns	Online backup	Backed up target	Disk unit's copy function	Application volume type (*1)	Required component (primary domain) (*2)
1	Not available	Application volumes	-	Any	GDS
2	Available	Slices detached temporarily	-	One of (*3): mirror (concat+mirror) (stripe+mirror)	GDS
3	Available	Proxy volumes	-	One of (*4): single mirror concat+mirror stripe+mirror	GDS GDS Snapshot
			Advanced Copy function	One of (*5): single mirror

(*1):

The table above describes the volume types according to the following classification.

Type	Description
single	Single volume created in a single disk.
mirror	Mirror volume connected to a mirror group to which one or more disks are connected. This is excluded when a lower level group is connected to the mirror group. Making disk data redundant can improve the services continuity.
concat	Volume created in a concatenation group. When no large volume exists, it can be created by concatenating multiple disks.
stripe	Stripe volume created in a stripe group. I/O loads of the services can be balanced up on multiple disks.
concat + mirror	Mirror volume created in a mirror group to which one or more concatenation groups are connected. This is excluded when a disk is connected to the mirror group. The effects of both concat and mirror can be gained.
stripe + mirror	Mirror volume created in a mirror group to which one or more stripe group are connected. This is excluded when a disk is connected to the mirror group. The effects of both stripe and mirror can be gained.
(concat + mirror)	Mirror volume created in a mirror group to which one disk and one or more concatenation groups are connected. Since the size of this volume is limited by the size of the disk connected to the mirror group, the effects of concat cannot be gained.
(stripe + mirror)	Mirror volume created in a mirror group to which one disk and one or more stripe group are connected. Since I/O loads to the disk connected to the mirror group are not balanced up, the effects of stripe cannot be gained. However, while the slice is temporarily detached from the disk, the effects of stripe can be gained since I/O loads to the volume are balanced up on multiple disks connected to the stripe group. Note, however, that a slice cannot be temporarily detached from a master volume related to a proxy volume.

(*2):

For an external backup server, GDS and GDS Snapshot must be implemented for creating shadow volumes.

(*3):

See "Creating a Snapshot by Slice Detachment."

(*4):

See "Exception to Proxy Configuration."

(*5):

See "Exception to Proxy Configuration" and "Using the Advanced Copy Function in a Proxy Configuration."

---

6.3.1 Backing Up and Restoring a Logical Volume with No Replication

6.3.2 Backing Up and Restoring through Snapshot by Slice Detachment

6.3.3 Backing Up and Restoring Using Snapshots from a Proxy Volume