Recently, our organization has seen an increase in calls from enterprises that have self-deployed Lync with challenges related to the front end (FE) pools. An initial comment such as “It was working fine the other day” is usually the first description of the problem that we are given. ”Our servers are up and running and all services seem to be fine”. So how does that relate to the FE pool?
There are a couple of common factors with the organizations that contact us:
1. They are running Lync 2013.
2. They are running a Lync 2013 Front End (FE) pool for High Availability (HA).
3. They recently have performed some type of maintenance or had an “event” with the FE pool.
After further investigation of these environments there is also one other common factor… they are running two servers in their Lync 2013 FE pool, and voila, here is the problem. But why you ask? Lync 2010 supported two servers in the FE for HA and it worked fine and you have probably even seen documentation that refers to having two servers in the Lync 2013 FE pool.
It is important to note that having two servers in a Lync 2013 FE pool does not automatically mean you will have problems. Some customers run this way (although not recommended) without issues for a long time. But it is when you have a server that goes down, or perform maintenance you may run into this issue, unless you bring the servers back online in a specific order or issue a cmdlet:
Reset-CsPoolRegistrarState -ResetType QuorumLossRecovery -PoolFQDN <FQDN>
This is done before bringing the pool back online as stated in the TechNet article found here, http://technet.microsoft.com/en-us/library/gg412996.aspx. The problem here is that this kind of defeats the purpose of High Availability. So what is the solution?
The solution is to have a minimum of three servers in your Lync 2013 FE pool as recommended in the same TechNet article referenced above. Why is the recommendation to have three servers in the Lync 2013 FE pool where we could have two in Lync 2010? This is because Microsoft has changed the architecture for Lync 2013 FE pools. In Lync 2013 user’s data is stored within routing groups, each routing group contains up to three Front End servers, a primary, a secondary and a tertiary. This essentially translates to the idea of a quorum for the Lync pools and there must be a minimum before the pool can start correctly. Without a quorum we see unpredictable behavior such as users having issues connecting. The routing group server assignment and quorum is controlled by a new Microsoft technology: Windows Fabric, this technology is used to create highly reliable, distributable and scalable applications .
The bottom line is if you are deploying Lync 2013 and you want to have High Availability, you deploy three Front End servers. If you want HA but have some capital our resource constraints, deploy a single server pool (non HA of course) and add the 2 servers when constraints can be addressed.