FT not supported on Nehalem systems with hyperthreading enabled?

[paulland]

I just got off the phone with vmware support and am perplexed.

I'm working on a brand new install of esx4 Advanced in a three-host datacenter and seeing some crazy behavior when trying to enable Fault Tolerance. Sometimes it works, but most of the time I wind up with an orphaned primary VM and a disabled secondary VM that is powered on but unmanageable. Or, it will just refuse to create the secondary, reporting that there are "insufficient system resources" (there are in fact plenty of resources available).

After a day of looking at configurations, shipping 500 meg worth of host logs, etc, my friend "Steve" in Bangalore tells me that he sees hyperthreading is enabled in the BIOS, and "we know FT doesn't work when hyperthreading is enabled."

"Steve" was unable to find any documentation on the public side of the support system but promised to find the relevant internal documentation and get it to me. But the systems did pass the Site Survey with hyperthreading enabled.

I'd think if this was the problem it would show up on the site survey (and would have been documented somewhere), but I'm new to vmware, so maybe that's not how it works here 😛

I haven't been able to find any evidence, either in the community or via Google, that anyone else is having this problem.

So, the question: Is "Steve" pulling my leg? Is anyone using FT successfully on a hyperthreading-enabled Nehalem host? If so, what did you do to get it to work?

Our configuration: 3 HP DL360g6 servers w/5530e processors, 24Gb, 6 NICs each.

TIA.

[AndreTheGiant]

and "we know FT doesn't work when hyperthreading is enabled."

I know a different thing.

There are a lot of "best practice" to make work fine FT:

- use similar speed CPU

- do not use HT

- ...

For HT the reason was that can create big delay if the same core it's also use by another VM. For this reason is best disable it.

Use Site Survey to check for other compatibility problem, and try to enable FT with the VM powered off... (This could resolv several problem, like disk type problem).

Andre

[depping]

Okay this is a myth that has been around for awhile and it's time to bust it. HT is fully supported for FT. And for Nehalem VMware actually recommends enabling HT as it can lead to improved performance. Don't think there's any supporting documentation right now that states that it is not supported. I will write a short article that it is supported so that you can at least find something on google...

Duncan

VMware Communities User Moderator | VCP | VCDX

-

---

Blogging:

Twitter: (*NEW*)

Available Soon: vSphere Quick Start Guide ()

If you find this information useful, please award points for "correct" or "helpful".

[AndreTheGiant]

And for Nehalem VMware actually recommends enabling HT as it can lead to improved performance.

If is true there there is some VMware documentation that must be update.

For example slide 10 of "Module 2: Using VMware Fault Tolerance" in "VMware HA and FT" couse under Partner University.

Andre

[depping]

Can you send me the link a PM? (And yes we are working on updating all material right now...)

Duncan

VMware Communities User Moderator | VCP | VCDX

-

---

Blogging:

Twitter: (*NEW*)

Available Soon: vSphere Quick Start Guide ()

If you find this information useful, please award points for "correct" or "helpful".

[paulland]

Well, it's good to hear that it's a myth, but VMware support doesn't seem to think so.

And if that's not the problem, support has no explanation -- yet-- of why FT doesn't work reliably in our environment. This is a very clean, brand new install on brand new equipment of ESX 4.0. No upgrades from 3.5, a garden-variety SAN, etc.

So, I'm still looking for answers. Who IS running 4.0 on this platform and using FT successfully?

Anyone?

Hello??

• crickets *

[paulland]

As mentioned before (see the OP) the site survey showed no compatibility issues.

As for the VM being off -- that's the only way we could even begin to make FT work; the process errors out if the VM is online.

I don't want to turn hyperthreading off on a hunch. I need all the performance I can get.... :smileymischief:

[depping]

Just to let everyone know the FT FAQ KB Article has been updated:

Does Fault Tolerance support Intel Hyper-Threading Technology?

Yes,

Fault Tolerance does support Intel Hyper-Technology on systems that

have it enabled. Enabling or disabling Hyper-Threading has no impact on

Fault Tolerance.

http://kb.vmware.com/kb/1013428

Duncan

VMware Communities User Moderator | VCP | VCDX

-

---

Blogging: http://www.yellow-bricks.com

Twitter: (*NEW*)

Available Soon: vSphere Quick Start Guide ()

If you find this information useful, please award points for "correct" or "helpful".

[admin]

Can you provide some additional details about what is happening? Perhaps send me email offline (kritter@vmware.com) since I don't know your details enough to look up your SR.

Thanks,

Karyn