Configuration needs for Fault Tolerance(FT)

[TMCishikawa]

Hello

I'd love to have your advice for FT configuration.

In " vSphere Availability Guide", it says "VMware recomends that you have a minimum of three host in the cluster". It also says its only "Recommendations".

in another day , i heard that VMware says " FT is supported even if its configurations are 2 ESXs in the cluster. ".

But i really can't imagine how works FT with only 2 ESXs.

in FT, primery esx send "logging traffic" to the secondary esx, then secondary esx do the job and send acks to the primery esx, and primery do the job finally

Once primery esx is failed, secondary esx become primery and third esx (spare one) become secondary. that I think how works FT is.

But if FT cluster has only 2 ESXs, after primery ESXs failed, the FT cluster has only one ESX(which is primery?). Primery ESX can do the job only after it receive acks.

but in this situation(FT cluster has only 2ESX and primery is failed), there is no secondary esx so that means primery can't receive acks and can't do the job.

Here is my questions, how can FT works if FT cluster has only 2 ESXs?????

[GixxerDave · Accepted solution]

The whole function of FT is to keep your VM up 100% of the time with out any down time by making a copy somewhere that is almost ready to start.

Now it does this by making a copy onto another host but is not turned "on". Like doing a vmotion on the VM but stops at 99% and then waits.

If you have only 2 host and one fails then once the 2nd host takes over the next step is to make a copy which it can't do because it does not have another host.

In a 3 host cluster

host 1 has live VM and host 2 has copy

host 1 dies then host 2 becomes live VM and then makes a FT copy onto host 3

It always wants to make a copy of the live host, so having 3 is recommended, but in a 2 host cluster it can't make the copy...and is no longer FT but your VM will still be running.

[TobiasKracht]

And if I will use shared storage for keeping my VMs? In this case even two node cluster will be HA

StarWind Software R&D