HP StorageWorks SVSP Virtualization Adapter for VMware Site Recovery Manager Application Notes (5697-0319, March 2010)

that are still running and shut down any designated VMs on the recovery site that were listed in the
recovery plan.
To run a recovery plan, select the Run button from the menu under the Recovery Plan window. SRA
takes the following steps when a recovery plan is run (a failover is issued):
1. SRA turns off the power to the protected site VMs.
2. On the recovery site HP SVSP storage, SRA will detach the replicated virtual disk and add host
permissions to the replicated virtual disks for the recovery site ESX host.
3. Site Recovery Manager will issue a host rescan of the storage adapters to detect the new virtual
disk mappings.
4. SRA uses the replicated LUN list provided as input to this command to map the recovery site
datastores to the recovery ESX hosts.
5. Starts the VMs using the failed over copies of the virtual disks.
NOTE:
Protection groups of one site always have the recovery plans that operate them on the other site.
After you run a failover, the failover is irreversible except by following the steps outlined in “Performing
a failback” on page 16.
Complete site failure
Creating application-level (transaction consistent) copies of the data is not a function of Site Recovery
Manager, but it is the most important point to address. The recovery point objective determines the
best way to re-establish consistency. HP SVSP asynchronous mirroring provides point-in-time copies
of the primary site virtual disks. If a complete primary site failure occurs, you must be prepared to
recover the applications manually on the DR site. This recovery can be relatively painless if the recovery
involves stateless applications such as web servers. However, recovery might be more involved with
recovery scripts running on system startup for DR VMs or with scripts running redo logs to regain
database consistency.
Performing a failback
Failback refers to the process of restoring applications and services back to the primary site after that
site is rebuilt or otherwise returned to normal operations. Currently, failback is not an automated
process. Future versions of Site Recovery Manager will provide this functionality. How a failback is
performed is determined by the type of failure that occurred originally at the primary site.
Follow these steps first to perform a failback of the VMs with HP SVSP:
1. Shutdown the VMs on the recovery site (if present) and remove them from inventory.
2. From the HP SVSP client, remove host permissions of the failed over LUNs from the recovery site
ESX host (previously the protected ESX host).
3. Rescan the storage adapters on the recovery site ESX hosts.
4. For each LUN containing a datastore of the protected VMs, create an asynchronous mirror group
from the acting protected site (the previous recovery site) to the acting recover site (the previous
protected site).
16