User Manual

Advanced Functions
5-19
• Mirrored write cache
A controller caches data in the memory for performance when the
delay-write option is turned on. To ensure that the cached data can be
retained, data written to one controller is also forwarded to the peer
controller. The RCC also serves as the path of data forwarding for
mirroring write cache. You may also disable the write cache mirroring by
the UI (see Mirrored Write Cache Control option on Section 2.7.15).
• Controller failover and failback
Both controllers monitor each other by periodic heartbeat signaling
packets exchanged over the RCC. If a controller detects that the RCC
link is offline or the heartbeat signal is not received within a period of time,
it will power off the peer controller and start the controller failover
procedure. The I/O access to the faulty controller will be redirected to the
surviving controller, and the background tasks will also be continued by
the surviving controller.
When the faulty controller is replaced by a new controller, the surviving
controller will negotiate with the replacement controller to perform the
controller failback procedure. The replacement controller will get
synchronized with the surviving controller to learn the latest status and
configurations. After that, the surviving controller will stop servicing the I/O
for the peer controller, and the replacement controller will take back all
the background tasks.
• Multi-path IO (MPIO) and controller preference
When using SAS host interface or using FC host interface without a switch,
the redundant-controller system requires MPIO driver installed at host side
for I/O redirection during controller failover and failback. You have to
symmetrically export the same LUN mappings for all host-interface ports,
such that a virtual disk can be seen by all I/O paths of both controllers
and controller failover/failback can work. After LUN configurations and
host-interface connections are done, please install the MPIO drivers
provided by your RAID system supplier to set up the redundant-controller
configurations.
The redundant-controller solution is compliant with T10 TPGS (Target Port
Group Standard) architecture. When a virtual disk is created, one of the
two controllers is assigned to it as the preferred controller (see the
Preferred Controller option in Section 2.6.2, 2.6. 4, and 2.6.5 for creating
JBOD disk, logical disk, and volume, respectively). When the I/O paths
from host computers to the virtual disk are available and the preferred
controller is online, the MPIO driver will dispatch the I/O to the paths of the
preferred controller for the virtual disk.