I just installed ESX 4.0 on a brand new Dell M610 blade with X5670 cpu's, and after installing and updating, and want to add it to my test cluster. No luck however, it turns out the machine somehow thinks it does not support EVC, and I cannot add it to my cluster with EVC enabled (with X5570 cpu's).
Ideas anyone? I already checked all bios settings, but saw nothing strange there..
By the way, I do have some other blades exactly like this one, also with X5670 cpu's, running fine with X5670 cpu's for another project.
VT and XD are enabled, see screenshot: http://files.hongens.nl/2010/06/05/m610_bios.png
Rebooted and disabled XD, then I saw an error on the console that Execute Disable was not enabled, still EVC not supported of course, then rebooted and enabled XD again.. No luck still.
Removed host from VC inventory and added again, still no luck.
This is strange by the way, cpuinfo result from two machines: http://files.hongens.nl/2010/06/05/cpuinfo.png. On the left is the problematic server, on the right a perfectly working server.. The problematic server is publishing sse3 and ss4 flags.. Euh? The right machine is a production machine, so I can't just go and change or check bios settings there..
Well, this tool, that VMotion info tool, shows the hardware should not be a problem: http://files.hongens.nl/2010/06/05/cpuinfo_tool.png (top one is problematic server)
Make sure the host is connected in VC and you can see thedmidecode command to verify all the hosts really are the same. also check /proc/cpuinfo on all the hoststo see if thecpuid and all the flags are the same for all your hosts. sometimes upgrading the bios can change the bios settings.
Verify all this
Hope that helps,
VCP2, VCP VI3, VCP vSphere
According to theVMware HCL when searching for EVC modes
the only Dell with 5600 cpus and supports EVC is this one:
Is this the one you have? I see here that it requires ESX 4.0 U1 and has some other patches so make sure those are installed too.
Hope that helps,
VCP2, VCP VI3, VCP vSphere
Yes, it's a 56XX cpu I have.. The article mentions patch ESX400-201002401-BG needs to be installed, which it is (I'm not too handy with VMware update, but once in a while I push updates using VMware update manager. On the new machines I installed all needed updates that were available.
Bulletin ID---- -
ESX400-200912403-SG 2010-06-06T18:37:53 Updates NSS and NSPR
ESX400-201005408-SG 2010-06-06T18:37:53 Updates bind-libs and bind-utils
ESX400-200912402-SG 2010-06-06T18:37:53 Updates OpenSSL
ESX400-201005404-SG 2010-06-06T18:37:53 Updates NTP
ESX400-201002404-SG 2010-06-06T18:37:53 Updates glib2
ESX400-201002406-SG 2010-06-06T18:37:53 Updates newt
ESX400-200912401-BG 2010-06-06T18:37:53 Updates vmx, vmkernel etc
ESX400-201003403-SG 2010-06-06T18:37:53 Updates acpid
ESX400-201002407-SG 2010-06-06T18:37:53 Updates nfs-utils
ESX400-201005403-SG 2010-06-06T18:37:53 Updates Expat
ESX400-201005401-SG 2010-06-06T18:37:53 Updates vmkernel64, scripts, tools etc
ESX400-Update01a 2010-06-06T18:37:53 VMware ESX 4.0 Complete Update 1
ESX400-201002402-BG 2010-06-06T18:37:53 Updates initscripts
ESX400-201005409-SG 2010-06-06T18:37:53 Updates sudo
ESX400-201003404-BG 2010-06-06T18:37:53 Updates lpfc820 driver
ESX400-201003401-BG 2010-06-06T18:37:53 Updates vmkernel64,vmx,hostd and scripts
ESX400-200912404-SG 2010-06-06T18:37:53 Updates DHCP
ESX400-201005407-SG 2010-06-06T18:37:53 Updates gcc packages
ESX400-201003407-BG 2010-06-06T18:37:53 Updates ehci-ehci-hcd driver
ESX400-201003406-BG 2010-06-06T18:37:53 Updates megaraid-sas driver
ESX400-201005402-SG 2010-06-06T18:37:53 Updates VMware-webCenter-esx
ESX400-201002401-BG 2010-06-06T18:37:53 Updates vmkernel64,vmx,hostd etc
ESX400-201002408-BG 2010-06-06T18:37:53 Updates Enic driver
ESX400-201003402-BG 2010-06-06T18:37:53 Updates hpsa driver
ESX400-201005406-SG 2010-06-06T18:37:53 Updates krb5 and pam_krb5
ESX400-201005405-SG 2010-06-06T18:37:53 Updates gzip
ESX400-201003405-SG 2010-06-06T18:37:53 Updates samba-client and samba-common
By the way, I installed another machine from the same batch as the problematic server, same problem..
So as an intermediate summary:
I have 4 blades, all exactly the same M610 blades with the same bios version, and the same bios settings.. Two are from a month back (EVC works), and two are newer, delivered last week. (EVC does not work)..
Only difference I have found so far is that the new two servers publish 3 more cpu features than the older ones: ssse3 sse4_1 sse4_2. Strange that both the old and new servers have cpu family 6 model 44 stepping 2. What's the difference? Is it Dell? Is it Intel?
I think I have to wait for VMware to integrate these new features in their update cycle (although some confirmation from a VMware rep to confirm this would be nice).
If moving VM's is your goal then the only way to do that in this situation is to power the VM off then Migrate (called a cold migration) this does not require CPU compatibility.
The EVC links I sent give a lot of caveats from VMware that not all cpu's or even the same processor families support EVC. If you dig down into all the links you might get more info but the reality is what you have found for now. If you have a lab environment and are really comfortable with this stuff you could play around with masking different cpu features (unsupported and not recommended by VMware so be warned) to try to figure it out. SSE4 would be my guess.
Other option is to have two sets of clusters. One with the older servers and the other with the newer to be able to migrate among themselves. To go between them you will need to power down the VM and make sure they all have shared Storage and network port names. Clusters are not required for VMotion only VMotion requirements.
Hope that helps,
BSCS, VCP2, VCP VI3, VCP vSphere
Yes. I have the same problem. getting pecuiar error. As per the the http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=100321... link the X5670 should support all type of EVC modes. But nos one is upporting that. Now i am thinking to upgrade the entire setup to Vcentre 4.1 and ESX 4.1 is it correct?
Now iam having two clusters because of this problem...