With the advancement of storage and Networking Technologies, we can see a huge amount of benefits in the way datacenter are being operated today, we look at some solutions like T2 Stretched Network, Storage Technologies like storage virtualization and replications between different sites to increase availability and minimize downtime.
Although we understand how both these technology operates. It has changed our understanding of how we want to provide availability & Disaster recovery to our datacenters in so many different ways and it has given us the option to have multiple choices in case of wanting to implement whatever kind of protection or high availability.
But this understanding has started to be misleading sometime where enterprises has thought for once that creating stretched datacenter with virtualization is an option that provides both choices in term of disaster recovery and high availability.
However, this idea is completely misleading and an absolute wrong due to the thinking of high availability at very high level without thinking about the software options and configuration that can be implemented.
Using VMware HA Feature has been one of the great use cases implemented by customers and has so many design consideration. I have heard people talking about vSphere HA design that tolerate server failure, Chassis Failure as well as a full rack failure. But I have not worked on designing a solution with stretched cluster to tolerate site failure with VMware HA.
When implementing stretched cluster for VMware Environment, you achieve high availability for sure. However, in case of a full site loose what does really happen.
The answer to this question is simple, all machines are going to be restarted on the other site but what kind of control as an administrator do you have. The only control you have is basically 2 options.
1- You can control is affinity/anti-affinity rules.
2- You can configure High availability restart priority. High, medium and low.
As per the above 2 option, I don’t see any kind of dependency mapping in case of failure for example, what about one web applications that require at least 3 different virtual machines that depend on each other being available. (Database up & Running, then Application server up and running, then finally web server up and running). I have not seen VMware vSphere options that can control restart priority in that way.