I'm new to Exchange database availability groups and Windows clustering in general. I've heard about the concept...
of "quorum." Could you explain what quorum is and why it's important to database availability groups?
A number of features within the Windows Server failover clustering model underpin each Exchange 2010 database availability group (DAG). Failover clusters employ a concept called "quorum." The failover cluster service for all DAG members must maintain quorum. If quorum is lost, the DAG terminates. This forces all the mounted databases within the DAG to dismount.
Consistency within a cluster is absolutely essential. Each member must always be consistent with other members of the DAG. The cluster database holds all the pertinent configuration information for cluster operations. If the cluster database cannot be loaded on a DAG member, the cluster service won't start. This happens because the cluster service isn't in a position to determine if the cluster member is consistent with the other nodes.
In a DAG with an even number of nodes, a "quorum witness" is used to avoid split-brain syndrome. Split-brain syndrome happens when multiple DAG member servers believe they own the same resources. The server that holds the quorum witness must be external to the mailbox database servers that participate in the DAG.
When quorum needs to be determined -- for example, when one or more nodes fail -- any member of the DAG that can talk to the witness server locks the witness server's cluster log file. The locked DAG member then receives an additional vote for quorum. All DAG members that are able to contact the locked node are in the majority, and thus maintain quorum. All others will lose quorum and cease to function as a part of the DAG.
Each DAG member gets a single vote, and the DAG member with the locked witness gets two. To maintain quorum, a majority of the voters (the DAG members plus the witness server) must be online and able to talk to each other. Failure to maintain quorum will cause the DAG to go offline.
DAGs with an odd number of members (starting from three, five and so on) use what is known as the "Node Majority" quorum mode. Each member of the database availability group is given a vote, and the majority of voters must be online and talking to each other. For example, if you have a DAG with five member servers, three of the five must be online for quorum to be maintained.
To summarize, quorum is having the majority of a DAG's nodes -- whether that's an odd number or an even number, plus the witness -- online and functional. If they are not, quorum is lost and the DAG ceases to function.
About the author:
Andy Grogan is an Exchange MVP based in the U.K. He has worked in the IT industry for the last 14 years -- primarily with Microsoft, HP and IBM technologies. His main passion is Exchange Server, but he also specializes in Active Directory, SQL Server and storage solutions. Grogan currently works for a large council in West London as the networks and operations manager supporting 6,000 customers on more than 240 sites. Visit Andy's website at www.telnetport25.com/.
Dig Deeper on Exchange Server setup and troubleshooting
Related Q&A from Andy Grogan
A coexistence period is an important part of any Exchange Server migration -- but why? Our expert explains. Continue Reading
Why is IMAP disabled in Exchange Server? When does it make sense to use it? Get the answers to those questions and more in this expert response. Continue Reading
An admin wants to know what his options are when it comes to checking Exchange 2010 server health. Our expert offers two suggestions. Continue Reading