A-17
Theory of Controller Operation
In either mode, arrays can only be accessed by the controller that
currently owns them. One controller will have no visibility to the
others’ arrays. If one controller fails, the surviving controller will
take ownership of all arrays. Pool spares and unassigned disks are
visible to both controllers.
The two controllers communicate via the back-end disk buses.
Although many different message types are sent between the two
controllers, the majority of the data is write cache data. In the
default write-back caching mode, any data that the host writes to
one controller is copied to the other controller before a good status
is returned to the host. By mirroring the data, if a controller fails,
the surviving controller will have a copy of all data that has not
been written to the disk.
Note:
The controllers communicate with each other using SCSI
initiator IDs 6 and 7 on each of the disk channels. You should
not change these IDs.
Each controller will send heartbeat messages to its partner
controller. If a controller does not receive a heartbeat message
within a set time period, it will kill the other controller assuming
that it has malfunctioned. After killing the other controller, it will
take ownership of the arrays and will activate its passive host port
to assume the defunct controller’s identity.
Table A-6 Presentation of LUNs in Active-Active Dual-Port Mode
Controller
Status
Controller 1
FC 1 Host Port
Controller 1
FC 2 Host Port
Controller 2
FC 1 Host Port
Controller 2
FC 2 Host Port
Both
controllers
online
Controller 1
LUNs
Controller 1
LUNs
Controller 2
LUNs
Controller 2
LUNs
Controller 1
fails
Inactive
Inactive
Controller 2
LUNs
Controller 1
LUNs
Controller 2
fails
Controller 1
LUNs
Controller 2
LUNs
Inactive
Inactive
DuraStor IUG - Online.book Page 17 Friday, July 19, 2002 3:06 PM
