Sun Microsystems StorEdge 3000 Series, Инструкция по установке, эксплуатации и обслуживанию

Страница: 36 / 318

1-10 Sun StorEdge 3000 Family Installation, Operation, and Service Manual • March 2004
1.5.3 Redundant Configuration Considerations
This section provides information about setting up redundant configurations for
increased reliability.
1.5.3.1 Host Bus Adapters
Fibre Channel is widely applied to storage configurations with topologies that aim
to avoid loss of data by component failure. As a rule, the connections between
source and target should be configured in redundant pairs.
The recommended host-side connection consists of two or more host bus adapters
(HBAs). Each HBA is used to configure a Fibre Channel loop between the host
computer and the array. In active-to-active redundant controller mode, the primary
loop serves the I/O traffic directed to the primary controller, and its pair loop serves
the I/O traffic to the secondary controller. The host-side management software
directs I/O traffic to the pair loop if one of the redundant loops fails.
1.5.3.2 Active-to-Active Redundant Controller
Since each fibre interface supports only a single loop ID, two HBAs are necessary for
the active-to-active redundant controller operation. Using two HBAs in each server
ensures continued operation even when a data path fails.
In active-to-active mode, the connection to each host adapter should be considered a
data path connecting the host to either the primary or the secondary controller. One
adapter should be configured to serve the primary controller and the other adapter
to serve the secondary controller. Each target ID on the host channels should be
assigned either a primary ID or a secondary ID. If one controller fails, the existing
controller can inherit the ID from its counterpart and activate the one standby
channel to serve host I/O.
1.5.3.3 Host Redundant Paths
The controller passively supports redundant fibre loops on the host side, provided
that the host has implemented software support for this feature.
In the unlikely event of controller failure, the standby channels on an existing
controller become an I/O route serving the host I/O originally directed to the active
channel on its pair of controllers. Moreover, application failover software should be
running on the host computer to control the transfer of I/O from one HBA to
another in case either data path fails.