Cluster wih some FT capable hosts, some not.....

[Pflynn]

Hello,

I have a need to build a new cluster with some existing hardware. Some hosts are FT capable, some are not. Within a given HA cluster, is it possible to have FT working if only 2 of 6 hosts or 4 of 8 hosts support FT? Because of the needs of the environment, I need to do n+1 for each cluster, so I am reluctant to create micro clusters of 2 or 3 hosts......

Thank you for any/all advice

Paul

[msemon1]

When you FT are you referring to the fault tolerant feature of ESX 4? If the answer is yes then you will need to create a cluster of ESX 4 servers for FT to work. You can't have ESX 3.5 and ESX 4 servers in same cluster with HA /DRS.

Mike

[weinstein5]

ESX 3.5 and 4.0 hosts can coexist in an HA/DRS cluster without issue - as long as none of the Vms have been upgradxed to version 7 of the virtual hardware -

If you find this or any other answer useful please consider awarding points by marking the answer correct or helpful

[Pflynn]

Hello, and thanks for chiming in

Sorry if it is unclear. I am indeed referring to the vsphere feature Fault Tolerance. All hosts are ESX 4, but it is existing hardware that is being repurposed. Some chips support FT and some do not.

My question is within a particluar cluster of hosts, can I mix FT-capable and FT-non-capable hosts, and make use of FT feature only on those hosts that support FT? I would not want to create separate clusters for only the FT capable hosts, as the end desire is to have n+1 per cluster,

Regards

Paul

[msemon1]

I don't think this will work. For FT to work your VM's will have to be on version 7 of virtual hardware. You are not going to want to mix ESX 3.5 and ESX 4 hosts with different virtual hardware versions with HA/DRS. Why not just create a cluster for your ESX 4 hosts and setup FT /HA/DRS for those VM's and hosts?

[J_Wood]

Good Afternoon Paul,

It seems the answer to your question is Yes, heterogeneous clusters are supported for Fault Tolerance however it is not recommended. CPU compatibility is not the only dependency here though, you have BIOS versions, patch levels and several other factors.

Take a look at the vSphere Availability Guide () for details on supported configurations as well at best practices for FT.

James Wood, VCP

Senior Systems Administrator | Arizona Department of Transportation | Phoenix, Arizona

[J_Wood]

One additional thought, how old is the hardware you are reusing? vSphere requires the CPU to support VT on the chip so if the older hardware is not x64 and VT compatible ESX 4 will not function on it.

James Wood, VCP

Senior Systems Administrator | Arizona Department of Transportation | Phoenix, Arizona

[Pflynn]

OK, thanks all for the response -

There is no 3.5 in the picture only 4.0 Enterprise Plus. The hardware is all relatively new, from a project that was discountinued before completion. I have just inherited servers that can be resued. All supoort VT. Not all support FT. There are two sites - I would am considering options w/regards to cluster creation. At each site, I need to maintain a swing server per cluster created (internal policy) and am wondering if I can have a single cluster with an EVC setting that will allow vmotions to work across all hosts, while having only a pair of servers that support FT within the cluster, so that I may host a limited number of FT-enabled vms

EX:

HW type CPU Processor CPU spee RAM Core type HT EVC mode FT

BL680 G5 4 E7340 2.4 Ghz 64 GB Tigerton 65 nm No Core 2 No

DL380 G5 2 X5470 3.33 Ghz 64 GB Harpertown 45 nm No Core 2 45 Yes

DL380 G6 2 X5550 2.66 Ghz 64 GB Nehalem i7 45 nm Yes Core i7 Yes

BL460 G6 2 X5560 2.8 Ghz 64 GB Nehalem i7 45 nm Yes Core i7 Yes

Best regards

Paul

[Pflynn]

OK, thanks all for the response -

There is no 3.5 in the picture only 4.0 Enterprise Plus. The hardware is all relatively new, from a project that was discountinued before completion. I have just inhereited servers that can be resued. All supoort VT. Not all support FT. There are two sites

EX:

HW type CPU Processor CPU spee RAM Core type HT EVC mode FT

BL680 G5 4 E7340 2.4 Ghz 64 GB Tigerton 65 nm No Core 2 No

BL680 G5 4 E7340 2.4 Ghz 64 GB Tigerton 65 nm No Core 2 No

DL380 G5 2 X5470 3.33 Ghz 64 GB Harpertown 45 nm No Core 2 45 Yes

DL360 G5 2 X5450 3.0 Ghz 64 GB Harpertown 45 nm No Core 2 45 Yes

DL380 G6 2 X5550 2.66 Ghz 64 GB Nehalem i7 45 nm Yes Core i7 Yes

DL380 G6 2 X5550 2.66 Ghz 64 GB Nehalem i7 45 nm Yes Core i7 Yes

BL680 G5 4 E7340 2.4 Ghz 64 GB Tigerton 65 nm No Core 2 No

BL680 G5 4 E7340 2.4 Ghz 64 GB Tigerton 65 nm No Core 2 No

BL680 G5 4 E7340 2.4 Ghz 64 GB Tigerton 65 nm No Core 2 No

BL680 G5 4 E7340 2.4 Ghz 64 GB Tigerton 65 nm No Core 2 No

BL460 G6 2 X5560 2.8 Ghz 64 GB Nehalem i7 45 nm Yes Core i7 Yes

DL380 G6 2 X5550 2.66 Ghz 64 GB Nehalem i7 45 nm Yes Core i7 Yes

DL360 G6 2 X5570 2.93 Ghz 32 GB Nehalem i7 45 nm Yes Core i7 Yes

DL360 G6 2 X5570 2.93 Ghz 32 GB Nehalem i7 45 nm Yes Core i7 Yes

24

[amvmware]

The way FT works is to ensure the VM is alwatys n+1, so if the first VM should suffer a failure it will automatically recreat another replica on another host - i would assume it would fail if it tried to recreate the vm on a host that was not ft capable as it assumes all the hosts are ft compatable in the cluster.

You also have the requirements for the logging nics to factor into the solution.