Awaiting too long for this!!! vCenter HA – Part 1
As of the GA of the vSphere 6.5, of course so do the VSAN 6.5, vROPS 6.4, LogInight 4.0 and vRealize Business for Cloud 7.2, it has been a bit busy for me to redeploying my lab environment. The pain point why I was not upgrading my existing 6.0 environment, it’s because there are no supported version of NSX being realised yet. So if you are currently using NSX, then vSphere 6.5 may not be a viable option for you yet. Actually Even for pure vSphere environment, you still need to ensure the interoperability and compatibility of Hardware and Software. Many of those are still yet certified in this stage, such as if you are upgrading the Site Recovery Manager 6.5, some SRAs are yet 6.5 supported.
Therefore, I think likely you would use vSphere 6.5 for development, test and evaluation purpose at this stage. Well, from my perspective, definitely I would like to test every what’s new as always. While you can refer to the VMware Official What’s New for vSphere 6.5, but in this blog I would like to talk about one most important item, vCenter Server HA.
It has been always a topic on how to protect the vCenter Server itself. If you are not too new to VMware, actually you can recall a tools named vCenter Server Heartbeat which is a separate product for protecting the vCenter. Honestly, I thought it was a great product, however, it has been discontinued. Well… I know many of you may ever suffered from split brain, unexpected downtime using the vCenter Server Heartbeat. But most of the cases I come across is because of configuration issue.
May be that’s normal for one not try treasuring something until we lost something. Before vCenter Heartbeat faded away, actually I saw few Customer looking into HA solutions for vCenter Server. But after it gone, more and more customers do wanna find better ways to protect their vCenter. Of course, it could be generated by the fact that vCenter becomes more and more important in Customer environment when newer solutions say VDI, Cloud… are being deployed and using with the vCenter. Even VMware, there is a new KB and white paper illustrating step by step configuration on how you should choose and deploy the protection on vCenter Servers thru’
- vSphere HA (High Availability)
- vSphere FT (Fault Tolerance)
- Native Watchdog
- Microsoft Failover Cluster
While it is trivial for vSphere HA deployment. vSphere FT requires quite high Requirement in network and limited sizing in CPU (Max 4 for SMP-FT in vSphere 6.0). Thus, previously I have tested on the Microsoft Failover Cluster before.
In the Version 6.5, here it comes another new vCenter Server Protection method which is named vCenter HA. And that’s why I have deployed a new lab and performed a vCenter HA setup to test out this cool new feature. In order to test this out truly for a Production environment, I have followed the following KBs and Documents in designing and deploying the setup:
- Supported and deprecated topologies for VMware vSphere 6.5 (2147672)
- Configuring Platform Service Controller HA in vSphere 6.5 (2147018)
- Configuring Windows PSC for High Availability in vSphere 6.5 (2147527)
- Configuring Netscaler Load Balancer to provide the PSC 6.5 High Availability (2147014)
- vSphere Availability Guide
So the first thing, we need to choose a deployment configuration best for your use case and environment. From the KB2147672, there are two configurations recommended. Recapping those as following:
Does not support Enhanced Linked Mode
Does not support Platform Service Controller replication
As the limitation of the configuration 1), I do not think it’s very legitimate to deploy such configuration in a large environment, where there are multiple vCenter Servers in an environment for performing cross vCenter vMotions or in the future the vMotion to the Cloud. And this is why I didn’t even consider to test the first option. However, originally I don’t want to deploy the second option too… as I need to setup more servers and load balancer for it. So I have consult VMware if my original design as following is supported.
Configuration 3) A mixed between Configuration 1 and 2 which putting PSC external but skip the dependence of a load balancer
What I got from Support, this configuration is also supported and no limitation on enhanced linked mode. Thus, if you do not have a load balancer, i would recommend you deploying this configuration. But do remember that this deployment method only protects the vCenter Server but not the PSC server. You PSC is the Single Point of Failure using this deployment. And this is why turn out I have actually deployed a Configuration 2 setup in the lab, but it takes some more steps in setting up the load balanced PSC. The detail steps will be blogged in separate posts.