Quorum Drive on cluster

[Arrosa]

I am having some queries about clustering. I know that the clustering utilizes quorum drive to duplicate the configuration level changes to the other node. In such case, I didn't get that what happens to a other (2nd) node active/passive cluster if the quorum drive is inaccessible or corrupted? Maybe I am having this doubt because I am not much good in clustering, specially when it comes to Quorum Drive. So please clear my doubt, and provide some more useful information about the Quorum Drive on cluster. Detailed information would be appreciative.

[Pratyush]

The role of quorum in a cluster is often misunderstood and can lead to the establishment of a configuration, especially with Windows Server 2008 where a new model is available. When a quorum is not reached, the deliberative body can not stand to vote and can not change the status quo. Thus, voters supported the status quo can block a decision in not calling a vote. The vote will then be automatically rejected and the status quo maintained. It is common to see called "quorum" disk used to store the cluster configuration. This is partly true. In a cluster configured with a quorum disk, this disk plays the role of arbiter.

[Sukhvinder]

Say in a cluster without a quorum, the quorum in a cluster represents all the voters to ensure the continued operation of cluster service and consistency this one. I later developed this aspect of continuity of operation. A nuance, however, with Windows Server 2008, a quorum model includes the quorum disk as voting. Regarding its usefulness, the quorum has three roles very simple:

• Provide a means to arbitrate the cluster membership - When starting a node, it uses a copy of this cluster configuration locally to identify the other cluster members. A mechanism then allows the sponsoring node during startup whether to join the cluster. If this server is sponsored by a node already started and if the cluster is functional, then retrieves a fresh copy of the cluster configuration (we call this action a JOIN).
• Help maintain the consistency of the cluster - This is the most important role of quorum. In a split-brain scenario, the quorum is used to ensure that shared resources are not erected on a single node. The quorum is used as a tie-breaker in this scenario.
• Provide a means of storing the cluster configuration - Within a cluster, each node must have a consistent view of the cluster configuration. This is made possible by the provision by a quorum of the cluster configuration stored in the quorum log.

[Pratyush]

The quorum log is a database containing the cluster configuration:

• The server part of the cluster
• Installed and shared resources
• The state of shared resources

The quorum log is stored by default in \ MSCS \ quolog.log. Suppose that the cluster is a human body and that the cluster members (nodes) are the lobes of the brain. Imagine that these lobes do not communicate. One can expect that the body no longer respond really well, orders from the two distinct lobes that are more synchronized, and thus leading to a catastrophe ... In the case of a cluster, a split-brain occurs when network links between two or more nodes will fail. The cluster is then split into one or more partitions that can no longer communicate.

[Solaris]

Avoid multiple nodes no longer able to "consult" perform operations on shared resources leads to an inconsistency in the service cluster. An Exchange cluster is housed in a 2 nodes cluster. Node 1 has the quorum node 2 holds the Exchange resources (databases).

• If node 2 has lost communication with node 1 while node 2 will attempt to take ownership of the quorum. If it fails, the cluster service on node 2 will stop itself because it will consider in poor health and unable to perform his role.
• Node 1 will attempt to recover the resources held by the node 2 and back in service.
• If node 2 is able to take ownership of the quorum, then it is that node 1 fails. Then node 1 which, if not already stopped or incapable of working, which will stop the cluster service.
• Finally, if either node 1 or node 2 does manage to take ownership of the quorum, the cluster is stopped on both nodes.

These operations aim to leave the clustered resource on a single node. You can not have two instances of the same service running on a network.

[Gefry]

In configurations and Majority Node and File Share Majority, the quorum configuration is stored on each cluster node on the system disk. The consistency of the replica (a quorum log) on all nodes is provided by the witness (the File Share Witness) or not the cluster itself, which determines a change is considered permanent as long as this configuration is present on the Half of the nodes. Two factors must be taken into account when selecting the model of a quorum:

• The presence or absence of shared storage in SAN or iSCSI / local cluster or geographic
• The number of failed nodes tolerated before total shutdown of the cluster

Take the case of an Exchange cluster nodes configured in the model and Disk Majority Node. Ultimately, the quorum configuration depends on a number of factors, the most important is to determine the number of failures tolerated. In practice, on clusters employing more than two nodes, an exit strategy must be defined to compensate for Murphy's Law. This strategy includes itself into a DRP (Disaster Recovery Plan) and most of the time, implies the presence of a cluster or at least relief procedures to reactivate the clustered application as soon as possible.