2.2.1 Virtual Machine Function

The virtual machine function is a function for the realization of virtual machine environment. This function enables you to create multiple independent virtual machines on one physical machine by virtualizing resources such as the CPU, memory, network and disk that are installed on a physical machine.

The function is made up of one host OS and multiple guest OSes.
PRIMECLUSTER makes it possible to build a cluster system between guest OSes on multiple host OSes or between host OSes.

Note

In a KVM environment, read the "host OS" as "hypervisor," in a VMware environment, read "host OS" as "ESX host."
When installing PRIMECLUSTER in a virtual machine environment, do not perform the following procedures:
- Temporary stopping the Guest OS
- Restart the Guest OS from a temporary stopped state
- Restart or stop the hypervisor (KVM environment), ESX hosts (VMware), without stopping the Guest OS

See

For details on the virtual machine function in a Xen environment, see "PRIMEQUEST Virtual Machine Function User's Manual."
For details on the virtual machine function in a KVM environment, see "Red Hat Enterprise Linux 6 Virtualization Getting Started Guide."
For details on the virtual machine function in a VMware environment, see the documentation for VMware.

Virtual machine function configuration

Cluster system in the virtual machine function

The virtual machine function provides the following methods to build a cluster system:

Building a cluster system between guest OSes on one host OS
Building a cluster system between guest OSes on multiple host OSes without the Host OS failover function
Building a cluster system between guest OSes on multiple host OSes with the Host OS failover function
Building a cluster system between host OSes

The table below shows uses and notes for each cluster system.

Method	Use	Note
Building a cluster system between guest OSes on one host OS	In a single physical server, build a cluster environment between guest OSes with the same cluster application configuration as that of the physical environment.	Since this cluster system is built with a single physical server, all cluster nodes will be shut down and operations will be suspended when the physical server fails. Therefore, this is not suitable to operate services. Do not perform the following procedures. Temporary stopping the Guest OS Restart from a temporary stopped state. Restart or stop the hypervisor without stopping the guest OS.
Building a cluster system between guest OSes on multiple host OSes without the Host OS failover function	To build a cluster environment between guest OSes with the same cluster application configuration as that of the physical environment and to use it as a development and test environment for cluster applications, or operate services	If the host OS fails, manual switchover of cluster applications is required because guest OSes cannot be forcibly shut down. Do not perform the following procedures. Temporary stopping the Guest OS Restart from a temporary stopped state. Restart or stop the hypervisor without stopping the guest OS.
Building a cluster system between guest OSes on multiple host OSes with the Host OS failover function	To build a cluster environment between guest OSes with the same cluster application configuration as that of the physical environment and operate services To switch the cluster application when the host OS fails	The ShutdownPriority attribute of RMS cannot be set on guest OSes. The active node may be stopped depending on the settings of survival priority. When a host OS failure is detected, the host OS is forcibly shut down. Then, all guest OSes on the failed host OS will stop. Do not perform the following procedures. Temporary stopping the Guest OS Restart from a temporary stopped state. Restart or stop the hypervisor without stopping the guest OS.
Building a cluster system between host OSes	To configure the following cluster environment and operate services: Monitors the states of both guest and host OSes and starts guest OSes on other host OSes if an error occurs on guest OSes or host OSes.	Hot-standby and scalable operations cannot be used. When a host OS failure is detected, the host OS is forcibly shut down. Then, all guest OSes on the failed host OS will stop.

Note

When installing PRIMECLUSTER in a KVM environment and building a cluster system between guest OSes, do not perform the following procedures.

Temporary stopping the Guest OS
Restart from a temporary stopped state.
Restart or stop the hypervisor without stopping the guest OS.

When building a cluster system between guest OSes on one host OS

This configuration allows you to run a cluster system on a physical machine. This is effective when you verify that the userApplication runs on PRIMECLUSTER.

When building a cluster system between guest OSes on multiple host OSes

This configuration allows you to continue work processes by a failover even if hardware such as a network or a disk fails.

Note

If the host OS cannot run, the node enters the LEFTCLUSTER state and manual switching by an operator may be required. For details, see "7.4.2 Corrective Action in the event of the LEFTCLUSTER state when the virtual machine function is used" or "7.2 Operating the PRIMECLUSTER System."

When building a cluster system between guests on multiple host OSes in a Xen environment or a KVM environment, you can use a function that automatically perform a failover when the host OS fails (Host OS failover function).

Host OS failover function

When building a cluster between guests in different units on a virtual machine, if an error occurs in the host OS, nodes in the cluster may enter the LEFTCLUSTER state. Host OS failover function allows for automatically switching cluster applications on the guest OSes in the case of the following errors in a cluster system between guests in different units in a Xen environment or a KVM environment.

- Panic of the host OS

- Hang-up of the host OS (slowdown)

This function is achieved by linking PRIMECLUSTER installed on the host OS with guest OSes.

Note that there are some precautions for operations, for example, setting the priority of RMS is not available by using this function. Then, you should take these precautions into consideration when designing the system.

Note

When creating a cluster application for a guest OS, do not set the ShutdownPriority attribute of RMS.
When a host OS failure is detected, the host OS is forcibly shut down. Then, all guest OSes on that host OS with a failure will stop regardless of whether they are clusters or not.
Do not build any cluster applications on the host OS.

Figure 2.1 Cluster system using the Host OS failover function on the virtual machine

Figure 2.2 Failover image in the case of host OS failure

Moreover, in a RHEL5-Xen environment on PRIMEQUEST, you can replicate the cluster system by doing live migration of guest OSes in which PRIMECLUSTER is installed or by copying the virtual machine image.

When building a cluster system between host OSes

This configuration allows you to continue work processes with starting up other guest OSes on other host OSes by a failover of the host OSes even if hardware such as a network or a disk fails.