5.0

Table 1-1. Differences Between Testing and Running a Recovery Plan (Continued)
Test a Recovery Plan Run a Recovery Plan
Effect on virtual machines at
recovery site
Local virtual machines are suspended
if required by the plan. Suspended
virtual machines are restarted after
the test is cleaned up.
Local virtual machines are suspended if
required by the plan.
Effect on replication Temporary snapshots of replicated
storage are created at the recovery
site. For array based replication, the
arrays are rescanned to discover
them.
In the case of a planned migration,
replicated datastores are synchronized,
then replication is stopped, and the target
devices at the recovery site are made
writable. During a disaster recovery, the
same steps are attempted, but if they do not
succeed, the errors are ignored.
Network If test networks are explicitly
assigned, recovered virtual machines
are connected to a test network. If
virtual machine network assignment
is auto, SRM assigns virtual machines
to temporary networks that are not
connected to any physical network.
Recovered virtual machines are connected
to a datacenter network.
Interruption Can be canceled. May be canceled in some cases.
How SRM Interacts with DPM and DRS During Recovery
Distributed Power Management (DPM) is a VMware feature that manages power consumption by ESX hosts.
Distributed Resource Scheduler (DRS) is a VMware facility that manages the assignment of virtual machines
to ESX hosts. DPM and DRS are not mandatory, but SRM supports both services and enabling them provides
certain benefits when using SRM.
SRM temporarily disables DPM for the cluster and ensures that all hosts in it are powered on before recovery
begins. After the recovery or test is complete, SRM re-enables DPM for the cluster, but the hosts in it are left
in the running state so that DPM can power them down as needed. SRM registers virtual machines across the
available ESX hosts in a round-robin order, to distribute the potential load as evenly as possible. SRM always
uses DRS placement to balance the load intelligently across hosts before it powers on recovered virtual
machines on the recovery site, even if DRS is disabled on the cluster. If DRS is enabled and in fully automatic
mode, DRS might move other virtual machines to further balance the load across the cluster while SRM is
powering on the recovered virtual machines, and DRS will continue to balance all virtual machines across the
cluster after SRM has powered on the recovered virtual machines.
Test Bubble Networks and Datacenter Networks
SRM can create a test bubble network to which recovered virtual machines are connected during a test. SRM
defaults to the Auto setting so that an accidental test recovery does not affect production. This network is
managed by its own virtual switch, and in most cases recovered virtual machines can use it without having to
change network properties such as IP address, gateway, and so on. A datacenter network, in contrast, is one
that typically supports existing virtual machines at the recovery site. To use it, recovered virtual machines
must conform to its network address availability rules. These virtual machines must use a network address
that can be served and routed by the network's switch, must be configured to use the correct gateway and DNS
host, and so on. Recovered virtual machines that use DHCP can connect to this network without additional
customization. Others require IP customization and recovery plan steps that apply the customization.
Virtual machines that must interact with each other should be failed over to the same test bubble network. For
example, if a Web server accesses information on a database, those virtual machines should fail over together
to the same network. This step enables testing of the function of the failed over virtual machines.
Chapter 1 Administering VMware vCenter Site Recovery Manager
VMware, Inc. 15