Networking considerations for multi-site clustering
The following tip is excerpted from a guest editorial by Symon Perriman, Senior Technical Evangelist at Microsoft who shared his expertise with us for an issue our WServerNews newsletter:
"There are several techniques used to reduce the time it takes for client to reconnect after a failover.
- Configure local failover first: It is recommended to have at least two nodes in every site and to configure the workloads to stay on the same subnet at the local site first before failing over to the secondary site. In a multi-subnet cluster this enables the client to remain connected using the same IP Address. Local failover can be configured using the preferred owner and possible owner properties for each cluster group. All the nodes at the primary site would be set as preferred owners and the nodes at the secondary site as possible owners.
- Reduce HostRecordTTL time: This cluster group property will adjust the time which a DNS record remains on a client's system before a new record is pulled down from the DNS server. By increasing the frequency that the DNS record is updated, the client will be able to see and access the new IP address sooner, allowing them to reconnect to the workload faster. By default, the record gets updated every 20 minutes, however in a multi-subnet cluster it is recommended to lower this value so that the record is replaced more frequently. For example, Exchange Server recommends a HostRecordTTL value of 5 minutes.
- Enable the RegisterAllProvidersIP setting: This cluster group property will adjust the information stored in DNS records for clustered workloads by including multiple additional IP Addresses which the workload can use (one for each subnet). Assuming that the client has the logic to try to connect to additional IP Addresses if the first address is unavailable, then this enables the client to reconnect much quicker in a cross-subnet failover. After a failover the client will first try to connect to the original IP Address they were using, and then they will try to connect to other IP Addresses in the DNS record and they should be successful."
To read Symon's entire guest editorial see Issue #933 Clustering, Replication, and Disaster Recovery of our WServerNews newsletter.
The above tip was previously published in an issue of WServerNews, a weekly newsletter from TechGenix that focuses on the administration, management and security of the Windows Server platform in particular and cloud solutions in general. Subscribe to WServerNews today by going to http://www.wservernews.com/subscribe.htm and join almost 100,000 other IT professionals around the world who read our newsletter!
Mitch Tulloch is an eleven-time recipient of the Microsoft Most Valuable Professional (MVP) award and a widely recognized expert on Windows Server and cloud computing technologies. Mitch is also Senior Editor of WServerNews. For more information about him see http://www.mtit.com.