Cisco Cisco Customer Voice Portal 8.0(1) Design Guide

Page of 223
 
4-17
Cisco Unified Customer Voice Portal (CVP) 8.x Solution Reference Network Design (SRND)
OL-15989-06
Chapter 4      Designing Unified CVP for High Availability
Gatekeeper
The gatekeepers share the same IP and MAC addresses. Therefore, if one of the gatekeepers fails, the 
hosts on the LAN are able to continue forwarding packets to a consistent IP and MAC address. The 
process of transferring the routing responsibilities from one device to another is transparent to the user. 
The H.323 endpoints (such as the Unified CVP H.323 Service, Cisco Unified Communications Manager, 
and gateways) register to a virtual IP address that represents the HSRP gatekeeper pair.
If one gatekeeper fails, its partner assumes primary control. The major disadvantage of HSRP is that both 
gatekeepers in the HSRP failover pair must reside on the same IP subnet or VLAN, therefore they 
generally cannot be separated geographically. Gatekeepers using HSRP for redundancy also do not share 
any state information. Therefore, when a failover occurs, all of the devices must re-register with the 
gatekeeper from scratch.
As of Unified CVP 3.1 SR1, HSRP is no longer recommended. Instead gatekeeper clustering and 
alternate gatekeeper configuration on Unified CVP is the preferred method of gatekeeper redundancy.
Gatekeeper Redundancy Using Alternate Gatekeeper
The Unified CVP H.323 Service can be configured with a list of alternate gatekeepers (as many as 
needed; there is no limit). When the H.323 Service starts, it attempts to register to the first gatekeeper in 
the list. If the registration is not successful, it tries the remaining gatekeepers sequentially in the list until 
a successful registration occurs.
The H.323 Service stays registered to that gatekeeper until either of the following events occurs:
  •
That gatekeeper has some type of failure. The H.323 Service recognizes a gatekeeper failure in the 
following ways:
  –
The periodic RAS Registration Request (RRQ) to the gatekeeper times out or is rejected.
  –
An Admission Request (ARQ) on a transfer times out.
  –
The gatekeeper pro-actively tells the H.323 Service to unregister, such as when the 
administrator does a shutdown on the gatekeeper configuration.
  •
The user does another setGK from VBAdmin. This causes the H.323 Service to register with the first 
gatekeeper in the list, if that gatekeeper is available; otherwise, it once again does a sequential 
attempt.
Gatekeeper clustering is not required in order to use Unified CVP alternate gatekeeper. It is possible to 
have two gatekeepers identically configured and also configure Unified CVP with alternate gatekeepers 
to provide redundancy.
The Unified CVP H.323 Service does not support gatekeeper clustering messages, but there is no reason 
that the gatekeepers cannot be part of a GUP cluster. In this way, other H.323 endpoints that natively 
support clustering (such as Cisco Unified Communications Manager and Cisco IOS gateways) can take 
advantage of the benefits of gatekeeper clustering. Unified CVP simply ignores clustering messages, 
such as when one of the gatekeepers in the cluster becomes overloaded or when Unified CVP registers 
with the gatekeeper.
Because Unified CVP does not automatically learn the other members of the gatekeeper cluster when it 
registers to the gatekeeper, it is necessary to define the gatekeeper cluster members statically in Unified 
CVP. Unified CVP uses one or more of the gatekeepers in the cluster as the alternate gatekeepers in its 
list and detects failure according to the rules mentioned earlier in this section.