Building a redundant Citrix environment requires a bit of thought and understanding of the different components. This is a high level review of what needs to be addressed when building High Availability into your XenDesktop 7.x design.
NetScaler Gateway. This is the first point of contact for remote users. If NetScaler is down, remote users will not be able to access and authenticate to the NetScaler Gateway which means no access for them. So at least dual NetScalers are needed to eliminate Single Point of Failure.
StoreFront. This is where the applications and desktops are enumerated for users that need to access them.Again redundancy is needed, at least a pair of StoreFront servers is needed.
Deliver Controller. Once users have successfully authenticated, StoreFront contacts the Delivery Controller to get a list of available resources (desktops/applications). So again N+1 redundancy is needed for the delivery controller.
Database Delivery Controllers require Database (Microsoft SQL). Now remember under the new FMA architecture, Local Host Cache feature is no longer availablein XenDesktop 7.x. So High Availability for the Database is required. Here you have a few options for High Availability. Citrix recommended option is mirroring. Why? Because the fail over is almost instant. With that said, Microsoft is planning on removing Mirroring from future SQL releases! There are other options for Database High Availability such as Clustering, High Availability on Hypervisor level, or AlwaysOn Availability Groups.
Provisioning Services (PVS) server. Delivery Controller have 2 connections available: Connections to the Hypervisor (MCS) and connections to PVS server. If using PVS and PVS server is not available when a client boots, then the users will not be able to access the content they need. A minimum of two PVS servers should be the minimum implemented per site.
Shared Storage. PVS needs shared storage to access boot image and vDisk. Now most likely that shared storage will be located on a highly available SAN connected to the PVS servers.
Hpervisor. You will need N+1 for redundancy on the infrastructure level as well. Also keep in mind if using MCS, those connections between the Delivery Controllers and Hypervisors will need to have redundancy built in as well.
License Server. You have one last Citrix component here which is the Citrix License Server. That is the only ccomponentyou do not need to have redundancy for. Why? Because it has 30 days grace period so even with license server down, the environment will still function. What happens if 30 days pass? If your environment is still down after 30 days, then frankly you have more serious issues to worry about!
So is that it?
Yes .. well almost. As far as Citrix components are concerned you have redundancy built in. However, there are other components you have to consider that Citrix environment depends on.
Domain Controller. Without a domain controller, your users will not be able to authenticate, therefore unable to access Citrix environment. N+1 is a given for your DCs.
DHCP server. This is needed for IP address assignment
DNS server. This is needed for name resolution.
This was a high level review of building redundancy into XenDesktop 7.x design
Thank you for reading.
Nick
Phill
November 28, 2014I did not know about Microsoft removing Mirroring in future!!! Typical Microsoft!!! What is the second best option?
Nick
December 13, 2014Hi Phil,
Good question. There are a few options:.
1) Windows Failover Clustering (WFC).A great solution that provides high availability. Downside is the failover process for large active enterprise database can take as much as 10 to 15 minutes which might be unacceptable for some organisations. Also setting up WFC is fairly complex and requires a specialized skill set that not all organisations might poses in house.
2) Hypervisor High Availability. Easy to set up but again similar to WFC, recovery can take take a few minutes while another virtual machine is started on another server. Also this option will not protect you against database corruption.
3) AlwaysOn.This is essentially the next generation in Database Mirroring that will be replacing the existing Database Mirroring! This requires Enterprise edition or higher and is pretty expensive and complex to setup. One thing to note with AlwaysOn is that Unlike WFC, AlwaysOn is a database-level technology and it does not provide the server-level protection traditional WFC does.
4) There are third party tools such Stratus Avance.This is software based that is installed on 2 nodes with a dedicated Ethernet connection. The software works through real-time data synchronization between the two nodes.This should provide you similar 99.999% High Availability as AlwaysOn.
5) Double-Take Availability is another 3rd party tool that can be utilized effectively to protect your SQL DB. It works by capturing and replicating byte-level changes to your disaster recovery server.
There are other 3rd paty solutions similar to ones mentioned above that will require an indepth post to cover them all.
I hope that gives you an idea of options available.
Nick
Jim
February 6, 2017Hello,
We have a planning scenario where we have a xenserver 6.5 or 6.7 running as a local host with the applications running to deliver virtual machines (Xendesktops as well as host a couple DC servers.
If we plan to move this current infrastructure model to two separate hosts with a SAN are the two xenservers setup in HA or is it something else that manages these systems?
Thank you