6

Our production environment has two redundant data centers. Within each data center we have multiple web servers talking to a single SQL cluster which has multiple broker DB servers.

The clustering works great if a single broker DB goes down, however, we have times that the cluster will need patching or fails and we want to auto fail over to a redundant cluster for the broker content without having to switch all web traffic over to the redundant data center.

Does anyone have something like this working in production?

I completely understand how to get the published data to the redundant cluster, but how would the web server traffic automatically pull from the passive DB cluster when the active DB cluster starts having problems?

Consider we are using 2013sp1 so the CIL and microservice architecture is not an option at this time.

  • 2
    I would expect there to be a load balancer between the web server and DB clusters, and to have the definition of what a "problem" is (e.g. Set of errors, timeout thresholds, etc) configurable (or preconfigured as part of the load balancer product default). Also, the master node in the cluster may have load balancing/failover capabilities. So configuring it point to the failover instance could be an option. This is more of a generic infrastructure/clustering configuration problem than Tridion specific. So I would seek out advice on infra/networking forums. – Nickoli Roussakov Feb 24 '17 at 13:57
  • @Nick, so have you seen scenarios in the Storage Config file where you point to a load balancer rather than a DB? – Chris Summers Feb 27 '17 at 11:22
  • I'm certain that I have, but I can't remember where. Sorry. But I do recall there is a different string used in cd_storage_conf for the JDBC config when pointing to a cluster. – Nickoli Roussakov Feb 27 '17 at 13:00
1

As Nick indicates, this isn't specifically a Tridion questions but SQL Clustering and Fail-over.

Perhaps one of these will be a good starting point?

https://msdn.microsoft.com/en-us/library/ms189134(v=sql.110).aspx (for SQL Server 2012)

https://msdn.microsoft.com/en-us/library/ms189134(v=sql.105).aspx (for SQL Server 2008 R2)

Also - I had a similar question in the past and the following was an example of a good piece of information regards mirroring (for single DB) and clustering (for SQL instace) failover... https://msdn.microsoft.com/en-us/library/ms191309(v=sql.110).aspx

1

We did use the NLB to connect to the Broker DB using F5 when we were using SQL 2008 back until 2015. Then we upgraded to SQL 2012 and removed the NLB and used the SQL failover.

From my experience of using it, I can say it works most of the times, but when you do have issues of connectivity, it becomes difficult to solve the same because the connectivity has an abstraction level w.r.t the NLB endpoint and how NLB is routing traffic to SQL servers. Most of the organizations have NLB configurations supported by operation teams which means to debug an issue you have to go through different teams which adds up time to find root cause and fix the same. So, please consider a risk-benefit analysis before going with an NLB approach.

Also, in this case the storage_config will point to the NLB end point and you have to setup and active passive or round robin on the NLB. We had an active passive setup between the servers.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.