Cisco Cisco Prime Infrastructure 3.0 문제 해결 가이드

The VMware ESXi vSphere High Availability (HA) and associated vSphere Distributed Resource
Scheduler (DRS) solutions provide a simple, efficient HA option for applications running in virtual
machines (VMs) in a virtual environment.  In the event of a failure in the physical host supporting
the VM, the affected VM is restarted on other server resources automatically, eliminating the need
for dedicated standby hardware and software.

The VMware HA and DRS solutions depend on a cluster, or group, of redundant hosts. When a
host is added to a cluster, the host's resources become part of the cluster's resources. The cluster
manages the resources of all hosts within it, and provides continued service when system
components fail. In the case of VMs, the cluster's admission control process (ACP) ensures that
the other hosts in the cluster can support the VM's processing and memory requirements in case
of failover. If the cluster does not support these requirements, the ACP will reject the VMs
admission to the cluster.

Cisco Prime Infrastructure (in both OVA and physical appliance forms) runs on VMware ESXi and
supports VMware HA. However, customers who want to use this feature must adjust the Prime
Infrastructure VM settings to do so.

As typically implemented, Cisco Prime Infrastructure running on a VMware ESXi VM reserves the
following host resources: 100% of CPU and 50% of RAM. For a Prime Infrastructure Standard
OVA with 16 vCPUs and 16GB RAM, this means that we reserve 16000MHz of the CPU and 8GB
of RAM for the VM's use (upon importation of the OVA).

But given that the Prime Infrastructure VM uses CPU and RAM reservation, the hosts in the
cluster, regardless of actual size, do not have sufficient unreserved RAM (plus overhead for
resource reservation) to support the Prime Infrastructure VM's admission to the cluster, so the
ACP will reject it. Of course, disabling the ACP would allow the Prime Infrastructure VM to be
admitted to the cluster. But this defeats the purpose of the ACP and HA itself, since there is now
no guaranteed availability of resources in case of the VM's host failure.

There are several approaches to ensuring that the Prime Infrastructure VM works with VMware
HA:

Reconfigure the Prime Infrastructure VM to use the vSphere option Percentage of cluster
resources option instead of the Host failures the cluster tolerates option. This requires
that the VM administrator know the size of the cluster and calculate the percentage of total

1.