Where does VMware HA and FT fall for me.

This is in response to a post that Duncan Epping posted a couple days ago which can be found here. It seems VMware wanted get some feedback on why customers were not virtualizing certain clustered software in vShpere or replacing cluster solutions with VMware HA. I don’t think it was defined well but I’d assume they only care about fail over clustering in this case.

Here is my take, from my perspective. The single biggest thing for me is that the current implementations of VMware HA and FT do not remove the operating system as the single point of failure. This is one of the reasons I have not actually implemented FT. Plus, FT has other limiting factors in its current form so I’d rather just stick with HA. This is not to say FT is useless and doesnt work because it does if you need a faster version of VMware HA. I know FT wasnt mentoined in the survey but it is a for of HA in my opinion.

I do run MSCS in vSphere but it’s more of a last resort option because of the added complexities. Clustering already has enough moving parts that could break in the system and make troubleshooting issues a pain without bringing virtualization into the fold. And typically, in the environments I’ve worked in, if clustering is warranted then the system is likely to be big enough and important to justify physical hardware for new systems. That’s not always the case but it really just depends on what fits the bill. We have database farms that are clustered on physical hardware which run multiple instances and still achieves consolidation. Last thing, the current clusters are not at the top of the list at the moment for virtualizing because they are running, working, and the hardware is already paid for.

Hopefully this customers perspective helps.

1 comment

Add yours
  1. udubplate

    Completely agree. I’m baffled by the number of vendors that don’t seem to understand that technologies such as FT do not rid of the SPOF a single operating system instance has inherently. It is a good fit for certain situations, but as you state, most applications/data that are configured in a failover cluster are like that for a reason due to their criticality. If there weren’t so many limitations, caveats, and complexities with failover clusters running on top of vSphere, there would be much greater adoption. Hopefully this improves in the near future as there hasn’t been much progression in the last couple years in this regard.

Comments are closed.