User`s guide

2.4 HSZ Array Controller Failover Operation

The HSZ array controller uses a transparent failover operation in which the

transfer of storage subsystem control occurs in a manner transparent to the host.

Two HSZ array controller modules installed in a StorageWorks controller

shelf and connected to the same host SCSI–2 bus operate as a redundant pair

with transparent failover. Figure 2–8 shows two controllers connected in this

manner. Each controller may be conﬁgured to have multiple SCSI target IDs.

A nonredundant (single) controller may have up to four IDs. A dual-redundant

controller conﬁguration may also have up to four IDs in any combination between

the two controllers.

In normal operation, each controller services only the targets it has been

assigned. In a failover situation, the surviving controller services all of the

targets from both controllers. Each target in any of these conﬁgurations supports

up to eight LUNs.

2.4.1 Transparent Controller Failover Resulting from a Fault

Once dual-redundant controllers establish a communications link in a normally

operating array controller storage subsystem, they maintain the link with periodic

status checks. If one controller fails, the other controller senses the situation and

begins the failover operation.

The sensing controller’s ﬁrst action is to assert a KILL signal, locking the failing

controller out of any further subsystem control. Once a controller asserts this

signal, it cannot be disabled by the companion controller. In the event that

both array controllers assert the KILL signal (as each senses a bad UART

connection, for example), the ﬁrst controller to assert the signal gains control of

the subsystem.

In normal operation, each controller keeps a record of the entire subsystem

conﬁguration. When a failure occurs, the surviving controller can then take over

control of its companion’s cache module and storage devices (this process is called

failover).

Note

RAIDsets, mirrorsets, and stripesets do not fail over for HSZ dual-

redundant controller conﬁgurations individually: all RAIDsets failover

together as the surviving controller takes over the relevant SCSI target

IDs and associated LUNs.

Note that the failover operation takes place at the controller level, in a manner

completely transparent to the host. As long as the redundant controllers reside

on the same SCSI–2 bus, the host is not required to be actively involved. The

transparent failover process is implemented in the HSZ array controller as a

simulated power fail situation.

2–20 Controller Technical Description