HA Eucalyptus requires the same requirements as non-HA Eucalyptus.
However, the infrastructure HA Eucalyptus will be deployed on must
meet some additional requirements, listed in the following
Redundant Physical Servers for Eucalyptus Components
Each cloud component (CLC and Walrus) and cluster component (CC,
SC, and VMware Broker) in an HA deployment has a redundant hot
backup. These redundant Eucalyptus components occur in pairs,
and each member of a pair must be mapped to a separated physical
server to ensure high availability.
HA pairs must be able to connect to each
If the HA deployment is to be able to tolerate the failure of
networking hardware, additional network interfaces are required
for the physical servers that host Eucalyptus components. The
physical servers hosting a CLC, Walrus, or CC and VMware Broker
must each have three network interface cards (NICs). Each
remaining physical server (except the NC components) requires
DNS Round-Robin Support
The DNS entries for the externally visible IP addresses of the
physical servers hosting CLC or Walrus components must be
configured to change round-robin style in an HA deployment.
HA Eucalyptus uses a kernel-level storage technology called DRBD
for storage integrity. DRBD must be configured to mirror data
operations between physical servers that host Walrus components.
For more information about DRBD, go to What is DRBD
For HA Storage Controllers, you must be using a supported SAN.
Only use HA SCs with NetApp or Equallogic drivers, not with the
iSCSI or JBOD SC driver.