Windows Server 2008 new arbitration /witness mode

Quorum Disk, also known as the quorum disk, is designed to coordinate failover between cluster nodes. In the Windows Server 2003 cluster, whether a single quorum device server cluster performs failover, quorum disk has the final say. This means that this is a way of dictatorship. Of course, for a single-node cluster, since there is no need for failover, it is not necessary to use quorum. As for the majority of node sets, although there is a single quorum corresponding to each node, quorum does not actually work.

So obviously, for a single quorum device server cluster, there is a single point of failure of the quorum disk, that is, if there is a problem with the quorum disk itself, the failover process cannot be coordinated, resulting in the entire cluster. There is no situation in the group. The improvement of this scenario has a new feature in the upcoming Windows Server 2008: Majority based cluster membership aims to give the node voting rights, avoiding possible quorum disk orders according to the principle of minority majority. Point the problem of failure, and avoid setting up the quorum disk as much as possible. In Windows Server 2008, you can choose the cluster mode according to the following principles:

l You don't need to avoid the quorum disk single point failure, use Disk Only mode. In this mode, only the quorum disk has voting ability, in fact, it is Previous single quorum device server cluster for Windows Server 2003

l To avoid single point of failure of quorum disk, use Node And Disk Majority mode. In this mode, quorum disk and each node have a vote

l To avoid the single point of failure of the shared data storage disk, the Node Majority mode is adopted. In this mode, each node has a separate storage device for storing shared data information, and the storage device is implemented on the hardware. Data synchronization, because in this mode, there is no need and no quorum disk, each vote can coordinate the failover process with its own vote. This scenario is mainly applied to the off-site disaster recovery plan of the banking and financial industry. >

l Need to support GeoClusters (nodes are clustered in different geographical locations) and the application requires shared data records In the high premise, for example, when designing the high reliability of DHCP Services, the Node And File Share Majority mode is adopted. This mode is to witness the shared folder on the server outside the node. This witnesses the shared folder and each node in the cluster. Hold a vote, in the event of a failure, through the voting mechanism to achieve failover