Feedback on VMWare FT for single DC, single network, and single SAN scenario

[drheim]

We have a single datacenter, with a single blade chassis, and a single SAN. If something happens to the datacenter, networking, SAN, or blade chassis our entire vSphere environment goes down.  Because of that, I really do not care for clustering with multiple VMs.  The extra complexity just typically does not make it work it in our environment, but I would love to hear opinions from people that have been working with VMWare extensively.  Our SAN does have deduplication across datastores, and trying to decide if VMWare FT would be good in our situation.  I guess if you have a blade crash(vmware host) because of hardware, you would not have to wait for HA to try and bring the server up on another host. Would you guys recommend using Vmware FT where possible in our environment? or one something like a fileserver would you just let HA take care of it if that one scenario of a hardware blade failure happens to occur.?  If VMWare FT works good and the replication is good enough to set it, and forget it let me know.  I really appreciate opinions from people that have been doing a lot of work in Vmware over the years. Thanks,

Dave

[daphnissov]

One thing to state openly here is that if you have all these single points of failure (SPoF), FT isn't going to help you. The only real solution to providing the higher availability that you apparently seek is to spend money by eliminating as many SPoFs as your budget will allow. FT is not a solution to replace vSphere HA and has very specific use cases but with some very serious caveats that impact how a vSphere environment works. And, that said, FT is most commonly used on systems that have no application-level availability options like a DC or file server. But again, this goes back to what your goals are and how to best achieve them. FT is not going to satisfy the goal of eliminating dependency on those items you mentioned.

[IRIX201110141]

We use FT in such  a environment. Customers with Dell VRTX (blade thing) have to move from Application based cluster to a easier manage solution and for us the VMware FT was the one and only solution based on a limit budget.

Please take in mind that FT 2.0 always creates a copy of all vDisk to another Datastore so that a replicated Storage isnt needed. Also VMware HA is always involved and needs to be enabled first.

Regards,

Joerg

[sk84]

And please note the necessary licensing (vSphere Standard, Enterprise or Enterprise Plus) and technical limitations:

Fault Tolerance Requirements, Limits, and Licensing

vSphere Features Not Supported with Fault Tolerance

By default, FT allows only 4 VMs and a maximum of 8 vCPUs per host. These values can be changed, but it is not a product to make a large number of VMs high available. And for vSphere 6.7 Standard and Enterprise licenses, only 2 vCPUs per fault tolerant VM are possible and for Enterprise Plus 8 vCPUs per VM (versions 6.5 and 6.0 have lower limits). In addition, a dedicated 10 Gbit network is recommended for FT which should probably not be a problem in a blade infrastructure.

More unpleasant is, that for FT VMs snapshots, storage vMotion, IO filters, storage policies and a few other vSphere features are no longer possible.

So, if it's a good idea to make your VMs high available with FT depends on whether you have the appropriate licenses and wether you can live with these limitations.