Vmotion between dual and quad core

[Aftenposten]

Hi all,

According to this pdf , vmotion is suppose to work between Intel 5150 dual core and Intel E5345 quad core on HP Blade servers. I tried this, but it only works when choosing low priority. Vmotioning from quad core to dual core times out and I get this error in vmkwaring on the source host: "Migrate: 6897: 4292461777: Unable to create scheduler group for migration worlds: VMotion failed to start due to lack of cpu or memory resources". Going from dual core to quad core I get the following error displayed: "A general system error occured: Failed to initialize migration at destination. Error bad00a0. VMotion failed to start due to lack of cpu or memory resources." Vmkwarning on the source host:

Oct 5 09:43:31 esx01 vmkernel: 0:20:21:15.700 cpu1:1133)WARNING: MigrateNet: 210: 4292461781: 0-0x3ea0db90:Received only 0 of 68 bytes: Migration protocol error

Oct 5 09:43:31 esx01 vmkernel: 0:20:21:15.700 cpu1:1133)WARNING: MigrateNet: 309: 4292461781: Start msg failed: Migration protocol error

Oct 5 09:43:31 esx01 vmkernel: 0:20:21:15.700 cpu1:1133)WARNING: Migrate: 1153: 4292461781: Failed: Migration protocol error (0xbad003e) @0x8c0c03

Server config:

Both run 3.0.2

esx host 1 is a HP BL480c dual core with 24 gb ram (production server pulled out for this test)

esx host 2 is a HP BL460c quad core with 6 gb ram

1 virtual server installed, Windows Server 2003

I have searche the forum for similar issues and tried some of the solutions others have tried, like masking, without success. Since HP claims this is supposed to work I'm not ready to give up just yet. Does the error message mean that vmotion between these to processors doesn't work or do they say that something else is wrong? In the latter case, what could that be? Any help you can provide will be greatly appreciated.

Thanks in advance.

Best regards,

Gaute

[psharpley]

No, the error message you have does not indicate a processor mis-match. Can you provide some details of your VMotion network? Is it dedicated or shared?

[nzsteve]

I'm with psharpley. I've never had problems with a vmotion between 51xx and 53xx family CPUs. AFAIK if it works on low priority it should work on high.

Check out your vmkernel vmotion network config. Sounds like the issue s more likly to be on that.

steve

[Rumple]

It sounds as if your SC is too buy to be able to do a vmotion on high. I would check esxtop to see if you are seeing high utilization on CPU0 and if so, check top in the Service console to see what is running alot.

Did you allocate maximum service console RAM and swap space when building the server?

[Aftenposten]

First, thanks to both of you for trying to help me.

Vmotion is shared. I've rechecked the vmotion configuration and it is identical on both servers (different ip addresses of course). I'm still able to vmotion to and from the one host I pulled out of our production cluster, to other dual core servers. The problem must lie within my test esx host. The network connectivity is identical for these two servers. In worst case I'll try to reinstall it. Please let me know if there is any specific information you would like me to post.

Thanks again.

Gaute

[Aftenposten]

Rumple,

Thanks for replying. I have allocated 800 mb to service console. I don't remember how much swap space I allocated, but it is the same on all my ESX servers. Any easy why to check swap space? Otherwise I would have to look it up in my notes.

I've never used esxtop or top so I need to read up on this utilities. Any tips you can give will be very appreciated.

Gaute

[psharpley]

Can you list the output of esxcfg-vSwitch -l?

When you say shared, what with? Does the VMkernel used for Vmotion have a GB link?

[Aftenposten]

Sorry about not being very precise. For reasons not neccessary to get into, these two servers only have 1 nic (GB link). I have enclosed the result of the command from both hosts. The one host I pulled out of production (host 1) has been running with this setup for over a year (we are to change this, but things sometimes take a long time) and I still can vmotion to and from host 1, only not to a quad core processor.

[psharpley]

Wow! one nic Well having Vmotion on it's own VLAN may help a bit. The error seems to indicate resource problems still - forget about dual and quad cores. Try running esxtop from the console and gives us that screen (q to quit out)

[Aftenposten]

Esxtop screen shots are included. The other 3 hosts we have in production have 4 nic's. Our first blade enclosure only has one ethernet switch, we are to replace that with Virtual Connect switches and probably add 2 more nic's to this host.

[Aftenposten]

Problem solved. Esx host 2 had some limitations set that made it impossible to use vmotion. After editing system resource allocation vmotion works like a charm. Thanks to everyone that replied and helped me with this problem.

Regards,

Gaute