Any EVC workaround available in this situation?

[pshearduk]

Hello,

I have a IBM Blade Center H chassis here at work, initially the chassis was poulated with 3 x IBM HS21 Blades, running ESX 3.5 and the Xeon processors are the X5355 family. All working fine and VMotion not a problem. Since then I have populated another 3 x HS21 blades in to the chassis as our infrastructure requirement has grown and these blades have the Xeon E5430 family processors inside them. These 6 blades are running approx 100 WinXP VM's and a Leostream connection broker (this environment is purely for VDI).

Now prior to VMware ESX 3.5 "update 2" I was unable to move and VM's from the 3 x Xeon X5355 family of blades to the new Xeon E5430 family of blades unless I did a "cold" VMotion. I decided simply because of the amount of WinXP VM's and my limited knowledge that I didnt want to attempt the CPU flag masking that I had read about here on the VMware forums and the VMTN. So I basically created two clusters each containg either the "old world" or the "new world" and decided that if I needed to free up some resources I would manually "cold" VMotion the VM's around as and when required.

Now when I read about the VMware ESX "update 2" and the EVC aspect of it, you can imagine that I was quite looking forward to finally being able to put both the "old world" and "new world" blades (Xeon 5355 and Xeon E5430) in to one HA/DRS cluster and have them all working together.. BUT guess what? EVC say's that my hosts are incompatible (sob)..

Can any experienced brains here suggest any other way I can get these VM's to VMotion accross the family of CPU's without supending or powering off the VM's?

yours hopefully

Paul

[Texiwill]

Hello,

Take 2 nodes one of each type and vMotion off the VMs. Then power them off and check your BIOS settings. EVC requires on HP Servers: Intel-VT/AMD-V be enabled as well as No-Execute Memory Protection be enabled within the BIOS. For the IBM blades I am sure there is something similar to enable. Once you do that then you can join these two nodes into a new VMware HA/DRS Cluster and enable EVC. Do this before you move any VMs onto the systems. Move a VM to one of the hosts then test migration between the two....

You may need to play with the BIOS until the settings are proper to enable EVC. Once you have EVC enabled migrate over the VMs and before you add a new node into the cluster modify its BIOS then join it. EVC should be enabled.

Best regards,

Edward L. Haletky

VMware Communities User Moderator

====

Author of the book 'VMWare ESX Server in the Enterprise: Planning and Securing Virtualization Servers', Copyright 2008 Pearson Education.

CIO Virtualization Blog: http://www.cio.com/blog/index/topic/168354

As well as the Virtualization Wiki at http://www.astroarch.com/wiki/index.php/Virtualization

[malaysiavm]

here is some article which guide to enable EVC in ESX with minimal impact.

http://malaysiavm.com/blog/how-to-enable-evc-on-esx-35/

Malaysia VMware Communities -

[pshearduk]

Thanks Edward, just before I read your post I was thinking exactly the same things.. so many thanks for underlining what I planned to do next!

I appreciate you taking the time to reply

kind regards

Paul

[pshearduk]

Thanks for the URL Malaysiavm, I will take a look

[Texiwill]

Hello,

Could you post the BIOS settings you need for the HS21? These would help the next person. I am also curios to see if this goes as expected. Looking forward to your followup.

Best regards,

Edward L. Haletky

VMware Communities User Moderator

====

Author of the book 'VMWare ESX Server in the Enterprise: Planning and Securing Virtualization Servers', Copyright 2008 Pearson Education.

CIO Virtualization Blog: http://www.cio.com/blog/index/topic/168354

As well as the Virtualization Wiki at http://www.astroarch.com/wiki/index.php/Virtualization

[pshearduk]

I'll certainly keep you and the group posted on all the details here

[pshearduk]

Edward,

I have a HS21 Blade that has ESX 3.5 Update 2 installed and we are awaiting a licence from VMware before we add to the HA/DRS/(EVC when working) cluster. So I am able to reboot / power off etc with no adverse effect on our environment.

I have just power cycled the blade and gone in to BIOS. This HS21 blade has the Xeon E5430 CPU's installed.

In BIOS I selected ADVANCED SETUP and then CPU OPTIONS. These are the options and how they are set as default :-

ACTIVE ENERGY MANAGER > UPPING ENABLED

PROCESSOR PERFORMANCE STATES > DISABLED

PROCESSOR ADJACENT SECTOR PREFETCH > DISABLED

PROCESSOR HARDWARE PREFETCHER > ENABLED

PROCESSOR EXECUTE DISABLE BIT > DISABLED

INTEL VIRTUALIZATION TECHNOLOGY > ENABLED

ENHANCED C1 POWER STATE > ENABLED

Can you see anything here Edward that I may need to ENABLE or DISABLE to get EVC working?

regards

Paul

[Texiwill]

Hello,

PROCESSOR EXECUTE DISABLE BIT > DISABLED

This most likely needs to be enabled.

INTEL VIRTUALIZATION TECHNOLOGY > ENABLED

This definitely should be enabled, which it is already.

Best regards,

Edward L. Haletky

VMware Communities User Moderator

====

Author of the book 'VMWare ESX Server in the Enterprise: Planning and Securing Virtualization Servers', Copyright 2008 Pearson Education.

CIO Virtualization Blog: http://www.cio.com/blog/index/topic/168354

As well as the Virtualization Wiki at http://www.astroarch.com/wiki/index.php/Virtualization

[pshearduk]

Hi Edward,

I'm now home from work, but I will certainly ENABLE the PROCESSOR EXECUTE DISABLE BIT on the IBM HS21 Blade tomorrow and let you and the group know what the outcome is. I have my VCP exam in a couple of weeks, and although I may not get questions about this kind of stuff, it's all invaluable knowledge.

Thanks again :smileygrin:

Paul

[mtakeuchi]

Just to put the information out there... Enabling "PROCESSOR EXECUTE DISABLE BIT" solved my problem of it reporting as an incompatible host.

My hosts are IBM HS21 blades w/ Xeon 5140.

[oldvbase]

I had the opposite problem, too much Execute Disable !. I tried the temp cluster idea and it started well. When I tried to VMotion from the 'old' to new 'new' cluster it failed compatability. I believe this was because of XD, which I suspect is masked away with EVC but present on the servers that are't hiding behind EVC.

any ideas anyone?

-

oldvbase

I used to be an Oracle, now I'm not really here