HP StorageWorks SAN Virtualization Services Platform Manager user guide (AG781-96017, March 2009)

Deleting async mirror tasks
Deleting an async task deletes the task, the destination virtual disk of the task, and the PiTs on the
destination virtual disk.
To delete an async mirror task:
1. Right-click the task.
2. Select Manage > Delete. A confirmation screen appears.
3. Click Finish. The task is deleted.
Synchronous mirroring
Synchronous mirroring helps you keep applications online in case of storage failures on the local site.
Synchronous mirroring allows you to have an online backup of a virtual disk on a different pool on
the same site so that you have a second up-to-date virtual disk available immediately if the first virtual
disk goes down. In synchronous mirroring, the host application receives a response to a write command
only after data has been successfully written to all mirror virtual disks.
Synchronous mirroring is performed by creating a sync mirror group to define the original virtual disk
to mirror. This virtual disk must already be presented to a host. The sync mirror group entity replaces
the original virtual disk as the access point for the host operating system. When you create a sync
mirror group, you automatically create the first task in the group. You can create PiTs and snapshots
on virtual disks that are in synchronous mirror tasks. A synchronous mirror task is an ongoing process
that directs the application’s I/Os to one virtual disk.
A sync mirror group operates in either of two modes:
• Always synchronized—If a task fails, all tasks in a group are paused. As soon as the agent that
is running the local mirror group identifies the first write error on one of the tasks, it starts rejecting
all subsequent I/Os, which informs the operating system that I/Os cannot be written.
• Continue on Fail— If a task fails, the application that is using the group remains active because
the VSM server uses the tasks which are still operational, if possible. This mode is the default mode.
The mirror service keeps track of any chunks of the virtual disk that were updated by some but not all
of the tasks in the group. If a task fails and recovers, the agent restores the chunks that need updating
on the recovered task from one of the normal tasks.
The mirror service features a mode called Dirty Regions, in which the agent (the fabric agent embedded
within the DPM) keeps track of the chunks that were most recently updated. Thus, in case of a recovery
from a state where one of the tasks was not up-to-date, and assuming other tasks were in a normal
status before the event, the agent copies the most recently updated chunks from one normal task’s
virtual disk to the virtual disks of the other normal tasks. This synchronization process is called Dirty
Progress. If Dirty Regions is not enabled, and the agent discovers that not all I/Os completed
successfully on all tasks, the agent considers the entire virtual disk as one region, and synchronizes
the normal tasks by copying the entire virtual disk from one task to the others. If there is only one task
in normal status at the time of the event, no such synchronization takes place, regardless of whether
Dirty Regions is enabled or not.
Running a group in Dirty Regions mode carries a small performance penalty that must be weighed
against the possibility of having to copy an entire virtual disk of data in order to restore a virtual disk
after failure. Dirty Regions is recommended for larger virtual disks.
Synchronous mirroring is also supported for cluster virtual disks. To set up a sync mirror group for a
cluster virtual disk, the virtual disk must be presented to a cluster of hosts before you create the sync
mirror group.
Using mirroring284