About High Availability in Sun xVM Ops Center
Sun xVM Ops Center provides high availability capability that enables you to manually transfer Sun xVM Ops Center Enterprise Controller functions from one system to another.
Overview
Sun xVM Ops Center high availability (HA) capabilities are based on transferring the /var/opt/sun/xvm directory structure from a primary Sun xVM Ops Center Enterprise Controller to a secondary Sun xVM Ops Center Enterprise Controller. The Sun xVM Ops Center HA configuration applies only to the Enterprise Controller and its co-located Proxy Controller. No HA configuration is available for standalone Proxy Controllers.
Most Sun xVM Ops Center data has been consolidated under the /var/opt/sun/xvm directory structure in version 2.1. A cron job on the Enterprise Controller periodically runs the habackup program to copy the small amount of Sun xVM Ops Center data that continues to reside outside of /var/opt/sun/xvm into a directory within /var/opt/sun/xvm.
In an initial HA configuration, the primary Enterprise Controller has Sun xVM Ops Center software installed, configured, and operational. The secondary Enterprise Controller has Sun xVM Ops Center software installed, but not configured, and not operational.
When the primary Enterprise Controller fails, the user initiates the failover to the secondary Enterprise Controller by:
- Shutting down the primary Enterprise Controller, if possible
- Preparing the secondary Enterprise Controller for failover
- Transferring the storage asset that holds the /var/opt/sun/xvm directory structure from the primary Enterprise Controller to the secondary Enterprise Controller
- Running the harestore program to restore the Sun xVM Ops Center configuration on the secondary Enterprise Controller
- Re-booting the secondary Enterprise Controller and starting Sun xVM Ops Center operations
The secondary Enterprise Controller takes over much of the primary Enterprise Controller's identity, including its host name, its IP addresses, its ssh keys, and its Sun xVM Ops Center data and role.
After the primary Enterprise Controller has been repaired, you can either establish it as a new secondary Enterprise Controller, or return operation to it by:
- Shutting down the secondary Enterprise Controller
- Preparing the primary Enterprise Controller for failover
- Transferring the storage asset that holds the /var/opt/sun/xvm directory structure from the secondary Enterprise Controller to the primary Enterprise Controller
- Running the harestore program to restore the Sun xVM Ops Center configuration on the primary Enterprise Controller
- Re-booting the primary Enterprise Controller and starting Sun xVM Ops Center operations
Caution – You must prevent the primary and secondary Enterprise Controllers from using the same IP addresses at the same time.
System Requirements
The systems that you use as Enterprise Controllers in an HA configuration must be identical in these characteristics:
- Processor class (SPARC or x86)
- Operating system (Solaris or RHEL 5.0)
- Sun xVM Ops Center software version, including updates
- Set of network interfaces that are cabled identically to the same subnets
- Use transportable storage to store the /var/opt/sun/xvm directory structure
The simplest solution is to use two systems of the same model that are configured identically.
The primary and secondary Enterprise Controllers must use identical network interfaces, that are cabled to the same subnets, so that the IP addresses used by the primary Enterprise Controller can be transferred to the secondary Enterprise Controller.
To use this HA capability, you must configure the /var/opt/sun/xvm directory structure on storage that you can readily move between the primary and secondary Enterprise Controllers. The best configuration would employ some sort of dual-ported or SAN-attached storage. The storage that you use to store the /var/opt/sun/xvm directory structure must offer performance that is sufficient to support Sun xVM Ops Center operation.
To use this HA capability, you must configure your storage device to hold the /var/opt/sun/xvm directory after you have installed the Solaris or RHEL 5.0 operating system, and before you install Sun xVM Ops Center software.
To avoid a single point of failure, the storage that you use to store the /var/opt/sun/xvm directory structure must offer its own data redundancy capabilities. Examples include detachable storage arrays that offer hardware data mirroring or similar data protection.
Consider all other single points of failure in your HA design, such as power, SAN, and network connectivity.
The primary and secondary Enterprise Controllers must use the same version of Sun xVM Ops Center software. You might not be able to verify the installed software versions at the time that you need to transfer Sun xVM Ops Center functions to the other system.
To verify the installed version of Sun xVM Ops Center software, view the content of the /n1gc-setup/.version.properties file. The product.version property lists the specific revision level of the installed Sun xVM Ops Center software. For example:
# cat /n1gc-setup/.version.properties #Note: This file is created at build time. #Mon Jan 26 08:49:53 PST 2009 version=GA product.version.xvmopscenter=2.0.0 date=2009/01/26 08\:49 product.version.xvmserver=1.0.0 build.variation=xvmopscenter product.version=2.0.0.820 #
Verify that the product.version property lists the same version on the primary and secondary Enterprise Controllers before you perform a failover procedure.
HA Technical Notes
- Only one Enterprise Controller, either primary or secondary, can be operational at any given time.
- It is possible for Sun xVM Ops Center user ID values to differ between the primary and secondary Enterprise Controllers. To address this, the habackup program backs up the local /etc/passwd file. The harestore program uses that information to change the ownership of the files to match the secondary Enterprise Controller's /etc/passwd file.
- User accounts and data that are not associated with Sun xVM Ops Center are not backed up or restored by this HA solution. Only Sun xVM Ops Center data is moved between the primary and secondary Enterprise Controllers.
- root user passwords on the primary and secondary Enterprise Controllers are not changed by this HA solution.
- UI sessions are lost on failover.
- The /var/opt/sun/xvm/osp/share/allstart directory is automatically NFS shared when you install Sun xVM Ops Center. If you use ZFS to provide the file system that mounts as /var/opt/sun/xvm, avoid using the ZFS sharenfs command to share /var/opt/sun/xvm/osp/share/allstart. This will allow the Sun xVM Ops Center software to use legacy NFS sharing tools to share the /var/opt/sun/xvm/osp/share/allstart directory.
- Use a user-defined tag to help identify the primary Enterprise Controller, and distinguish it from the secondary Enterprise Controller. You can add a tag by using Edit Gear, after you select an Enterprise Controller from the All Managed Gear list. See the following procedures for more information about adding user-defined tags:
- You must install, but must not configure, Sun xVM Ops Center software on the secondary Enterprise Controller. The data that is saved by the habackup program on the primary Enterprise Controller is used by the harestore program on the secondary Enterprise Controller to duplicate how the primary Enterprise Controller is configured. See the following procedures for more information about Sun xVM Ops Center software installation:
Where to Go From Here
See Configuring Storage for High Availability for a simple example of configuring storage to support Sun xVM Ops Center high availability.