About Clusters |
A cluster consists of a pair of storage systems that are connected through an interconnect adapter and cables, and are configured so that both storage systems share access to a set of Fibre Channel disks, subnets, and tape drives to provide fault tolerance.
When one storage system in a cluster undergoes a system failure and cannot reboot, the partner storage system in the cluster can take over the failed system's functions and serve network clients the data on the failed system's disks. This is known as a takeover. In addition to an automatic takeover, an administrator can issue a manual takeover at any time to force a takeover, for instance, to allow scheduled maintenance to be performed on a storage system.
An emulated storage system is created on the takeover system. This is a software copy of the failed storage system that is hosted by the takeover storage system. The emulated storage system appears to users and administrators like a functional version of the failed storage system.
After the failed partner is running normally again, you issue a giveback, which returns the identity from the emulated storage system to the failed system, resulting in a return to normal operation.
The storage systems in a cluster have the following characteristics:
They are connected to each other through a cluster interconnect adapter and cable through which they
Continually check whether the other storage system is functioning
Mirror log data each other's NVRAM
Synchronize each other's time
They use two Fibre Channel Arbitrated Loops (FC-AL) through which each storage system
Uses the A loop to manage its own disks
Uses the B loop in takeover mode to manage its partner's disks
They each own their spare disks and do not share them with the other storage system.
They may reside on the same NT domain or on different domains.
If one storage system in a cluster stops functioning physically, its partner storage system accomplishes the following operations:
Creates an emulated storage system using the failed storage system's identity
Accesses the failed storage system's disks and serves its data to clients
Maintains its own identity and continues to serve its own data to clients
When the failed storage system is functioning again, the following operations occur:
An administrator initiates a giveback that terminates the emulated storage system on the partner.
The failed storage system resumes normal operation, serving its own data.
The cluster resumes normal operation, with each storage system ready to take over for its partner if the partner fails.
During normal operation, in addition to serving data, the cluster performs the activities described in the following table.
|
Action |
Description |
|
Heartbeat transmission |
The heartbeat signals transmitted through the cluster interconnect adapter and cables tell each storage system the state of the other storage system. This information is written to disks on each storage system that the cluster monitor designates as mailbox disks. The disks provide redundancy. |
|
Monitoring |
The cluster monitor software looks at the heartbeat signal, commands from storage systems, and disk connections, and initiates a takeover under the following conditions:
Note that if the storage system that is not transmitting a heartbeat is rebooting, no takeover occurs. Takeover might be disabled if NVRAM mirroring is not functioning. Automatic takeover is disabled if the partner mailbox disks are not accessible. |
|
NVRAM mirroring |
Each storage system dedicates half of its NVRAM to a synchronously updated copy of its partner's NVRAM. If a takeover occurs, the takeover storage system uses the cluster data in the part of the NVRAM dedicated to the failed storage system to ensure that no information is lost. |
|
Time synchronization |
Filers' times can deviate from each other. A daemon updates the times on storage systems periodically to keep them synchronized. |