I am interested in the two features named "Fault Tolerance" and "App High Availability" which are part of some VMware products. As far as I understand the vCenter Server instance is needed to perform administrative tasks. Therefore, a crash of the vCenter Server makes it impossible to modify certain aspects of the VMs, but the VMs still run.

I also found out that the "HA" feature still works (with some minor limitations) even when no vCenter Server is running.

Now, my questions:

What happens to a VM which is configured to use FT if the vCenter Server is offline and the host of the primary VM is offline? Ideally, FT would take care to seamlessly continue the VMs operation on the secondary host.

What happens to a VM which is configured to use App HA if the vCenter Server is offline and the app inside the VM crashes? Ideally, the app crash would be detected and depending on the configuration the app or the whole VM would be restarted.


Some of this is covered in the vSphere HA Technical Deep-Dive...

But to summarize, vSphere installs HA agents on the hosts when you enable HA. vCenter is not required for HA to function. vCenter is needed in order to make changes to the cluster configuration, however, HA restarts follow a set prioritization schedule.

  • In your VMware Fault-Tolerant (FT) example, the secondary would take over.
  • vSphere App HA would restart the application or VM (assuming a supported application).

The point is that these things HAVE to be able to work without vCenter. E.g. the situation in which the host containing the vCenter fails needs to be covered (since it's highly-probable), and is.

  • 1
    @C-Otto Yes... but I would really look at handling application failover using standard tools/clustering before considering App HA. And VMware FT isn't that great either... Far too many limitations for it to make sense. – ewwhite Jul 31 '14 at 12:01
  • I am in favor of Pacemaker/Heartbeat/Corosync/DRBD, but I need to show other options. Regarding FT: Application crashes are not detected, and SMP is not supported. Other than that I do not see major problems. Regarding App HA: This is very expensive, and I fear that my applications are not supported out of the box. Furthermore, a full VM reboot might take too long. – C-Otto Jul 31 '14 at 12:04
  • FT not being able to use SMP is a deal breaker in almost every case. For App HA, the cost isn't worth it, as things can often be handled at the application level and by leveraging anti-affinity rules to separate VMs. I can tell you that in my environments, we just take the 30-60 second HA outage. But then again, hosts just don't fail like that without a reason... – ewwhite Jul 31 '14 at 12:16

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.