Server switchover is a feature that allows the user to switch over applications from a primary server to a predefined spare server when the primary server fails or is stopped for maintenance.
It can also perform Auto-Recovery, which automatically switches applications over to a spare server when a hardware failure is detected on the primary server.
Server switchover can be realized through two different methods, as described below. The switchover method used by a given server is automatically selected when configuring its recovery settings. If HBA address rename settings were already made when applying recovery settings, the HBA address rename method will be automatically selected. Otherwise, the backup and restore method will be selected.
However, each method has its own restrictions regarding the supported hardware environment. Refer to the "Note" in "1.2 Hardware Environment" of the "ServerView Resource Coordinator VE Installation Guide" for details on the hardware environments supported by each switchover method.
Backup and restore method
In a local boot environment, this method restores a system image backup to a spare server, which is then automatically started up. This method is used when no HBA address rename settings have been made to the primary server.
After switchover, the operating system and its applications will resume on the spare server from the state they were in at the last system image backup. Note that only the content of the first local disk (or boot disk) as seen by the BIOS of the managed server is subject to a backup or restore operation, including all partitions (Windows drives or Linux partitions) present on the boot disk. However, since additional disks (disks used as data disks), are not subject to backup and restore, their content can not be made accessible to the spare server with a server switchover. When using more than one local disk, backup and restore of such additional data disks should be performed using external backup software.
HBA address rename method
In a SAN boot environment, this method sets the WWN of a spare server's HBA to the same value as that originally set on the primary server. This allows the spare server to connect to and boot from the same boot disk that was used by the primary server. This method is used when HBA address rename settings have been made on the primary server. Because this method automatically starts the spare server from the primary server's boot disk, applications can be resumed without users being aware of the hardware replacement that occurred.
The HBA address rename method is also referred to as an I/O virtualization method.
For PRIMERGY BX servers, the network configuration (VLAN IDs of adjacent LAN switch ports) of the primary server will be inherited by the spare server. This is true for all methods of server recovery.
After two servers have been switched over, a failback operation can switch them back to their original configuration using the same switchover method. Conversely, server takeover is the operation that appoints the activated spare server as the new primary server instead of switching back servers to their original configuration.
Note
Auto-Recovery occurs when a hardware failure is detected. However, it does not occur when the operating system has stopped as a result of a software error or when the operating system is automatically rebooted.
Refer to "9.3 Server Switchover Conditions" for details on server switchovers.
Furthermore, since the Auto-Recovery is driven by a hardware failure, it prioritizes a fast recovery to a spare server instead of collecting an OS memory dump (which would be used to troubleshoot an OS failure). Thus, even if a memory dump was set to be collected in the event of an OS failure, server recovery will take precedence and the memory dump will not be collected.
Server switchover can be realized by either one of the following methods: Backup and restore method or HBA address rename method.
When configuring the HBA address rename function as the switchover method, first confirm that the HBA address rename settings have been configured properly before configuring the server switchover settings.
It is not possible to specify spare servers for individual VM guests. It is either possible to store VM guests on a SAN or NAS shared disk and assign a spare server to the VM host, or use the VM high-availability feature provided by the server virtualization software used.
Refer to "A.3 Functional Differences between Products" for details on the different high-availability features available in each server virtualization product.
For individual VM hosts, server switchover can be realized either from Resource Coordinator VE or from the high-availability feature provided with the server virtualization software used.
When using HBA address rename, a spare server can be shared between servers running physical OS's and those running VM hosts by combining the following settings.
Refer to "Figure 9.3 Sharing a spare server between physical OS's and VM guests" in "9.2 Configuration" for details.
Specify a VM host that doesn't run any VM guests as the spare server used to recover VM guests within the high-availability feature of the server virtualization software used.
Specify the physical server running the above VM host as the spare server of other physical OS's.
Once the above settings have been made, a server failure will trigger the following recovery operations. If the failed server was running a physical OS, Resource Coordinator VE will shut down the VM host on the spare server and switch the failed physical OS over to the spare server. If the failed server was a VM host running VM guests, the high-availability feature provided with the server virtualization software will recover the VM guests on the spare VM host server. Since physical OS's and VM guests share a common spare server, the two types of recovery described above can perform together: once one type of recovery occurs on a spare server, the other type of recovery can no longer be performed on that same spare server.
Information
Server switchover based on backup and restore takes approximately 3 minutes, plus the time required to restore the system image. Image restoration time depends on different factors such as disk space and network usage, but as an estimate, a disk of 73GB will take 30 to 40 minutes (the transfer of the system images takes between 10 to 20 minutes, while system restarts and other configuration changes take another 20 minutes).
Server switchover based on HBA address rename takes approximately 5 minutes, plus the time required to start up the original operating system and services on the spare server. If a VM host was running on the spare server, the time required to shutdown the spare server must also be included.
