FT: reduced cpu speed when activated

[jonathan_rober1]

Hi everyone, I have a strange situation in my cluster.

I have a 2 node cluster with HA activated. (esx6)

I have a few VM and one of them is FT protected.

I have a 10Gb connexion between the 2 node for FT and another one for VMotion.

I am working on the FT protected VM, I perform a task which take about 1min and 15s when not protected with FT.

I enable FT (with the same number of vCPU of course) and the same task take about 3min to be performed. I had a look to FT bandwith on 10Gb link and it is ok (le flow is not really high).

We tried to stress CPU whit CPUZ to 100% for some time (maybe not a good idea) and the CPU capacity decrease in time really fast (from near normal to really slugish and non usable VM). At the same time, the flow on 10Gb link increase continously, the sluggier the VM goes, the more flow we get on the link.

We have tried a lot of thing (change cable for 10GBs, patch network card...) and we have no more ideas now.

If you have some, that would be great.

By the way, we have tried to use legacy FT on our cluster. It works perfectly, we are limited on 1vCPU but we can use CPUz without getting the sluggishness appearing. And the test job is performed in 2 minutes (which is normal with only 1 vCPU).

[vbrowncoat]

There is an overhead to using FT, that is what you are experiencing. This can impact all aspects of a VMs performance. You might see if you can add additional CPU (if you aren't already at 4) or more memory to see if that will improve performance.

Understand that any changes made on one VM have to be made at the same time on the secondary VM (they have to be acknowledged on the secondary before the primary can move on), which is why when you ran your "stress test" bandwidth usage increase and performance decreased.