Redundant Host API
Software connection is the inverse action to software disconnection: it starts the drivers for PCI devices in the domain and resumes normal operation. When initiated for a domain in the DISCONNECTED state, it brings the domain to CONNECTED state through the intermediate CONNECTING state. Software connection can be used to cancel the effect of software disconnection for a domain during switchover preparation. For example, suppose that two domains should be switched over simultaneously in an atomic transaction; software disconnection succeeded for the first domain but was rejected for the second domain. As a result, the switchover is not possible and the first domain should be brought back into operation by software connection.
The same states apply to separate slots in the domain. They can be retrieved on a
The requesting host may specify a timeout for software disconnection. This value, expressed in milliseconds, serves as an indication to the owning host of the time interval during which the software disconnection should be completed. The requesting host indicates that after the expiration of the timeout it intends to either abandon the switchover or perform forced switchover.
After the software disconnection of the relevant domains is complete, switchover is initiated to change ownership of the domains. To trigger the switchover, the RhPerformSwitchover function should be called.
After the switchover, software connection is automatically initiated for the relevant domains on the receiving hosts. It is not necessary to call any functions after the switchover to software connect the received domains.
6.2.5.1.2Partially Cooperative Switchover
With this type of the switchover, software disconnection takes place for some but not all of the devices in the domain. It may be considered that some devices need to be prepared for switchover while other devices may be switched over without preparation.
Another possible scenario is that some devices are considered “more important” and the others “less important”. The switchover is initiated as soon as software disconnection completes for “more important” devices, without waiting for completion of preparation for “less important” devices.
In all these cases, at the moment of switchover some devices are prepared for switchover, while other devices are not and may need to be brought into a known initial state after the switchover.
After the switchover, software connection is automatically initiated for the relevant domains on the receiving hosts; so it is not necessary to call any functions after the switchover to software connect the received domains.
6.2.5.1.3Forced Switchover
In the forced switchover scenario, the domains are not software disconnected before the switchover, so device operation is not quiesced and for the device drivers and other software on the resigning host the PCI devices physically disappear, possibly in the middle of transactions. PCI devices are generally in an unknown state after the switchover. However, if the parameter Reset is used in the RhPerformSwitchover function, the PCI buses of the domain are reset, which brings the devices into the known initial state on the new owner host.
62 | High Availability Software for the Intel® NetStructureTM ZT 4901 Technical Product Specification |
