ESXi upgrade from 3.5 to 4.0 error: NUMA node 1 has no memory

[meetom]

I was running ESXi 3.5 and decided to upgrade to 4.0. During the upgrade process, the console gave me this error message: "The BIOS reports that NUMA node 1 has no memory. This is either caused by a bad BIOS or a very unbalanced distribution of memory modules." See attached screen shot.

Here's what I have tried so far.

1. swap all the memory modules out with new ones and got the same error

2. did a install instead of an upgrade and got the same error

Please keep in mind it was working on ESXi 3.5. Any ideas?

[Troy_Clavell]

maybe this KB article will offer some guidance.

http://kb.vmware.com/kb/1003690

[testqa]

Hi

May I know the server model on which you are trying to upgrade?

If you fully populate the memory banks then I think you will not hit this error.I think the server has more than one CPU and the memory is not distributed.

Thanks

[meetom]

I'm running a whitebox machine with the following configs

- Tyan Thunder h1000E (S3970G2NR) manual: http://www.tyan.com/manuals/m_s3970_110.pdf

- 2 dual core AMD CPUs

- fully propulated memory in all 8 slots (16GB total)

I have no problem running ESXi 3.5. Upgrading to ESXi 4.0 has this issue.

[meetom]

I actually swapped out the CPUs to 2 quad cores AMD (Shangai) CPUs and swap out all the memory modules. Again, the NUMA error came up again. If anyone has any ideas then that'll be great.

BTW, SRAT is enabled in the BIOS.

[esxtek]

Did the hardware diagnostics run fine.

[meetom]

This time I completely isolated each hardware component and install ESXi. The hardware is good, but ESXi fails every time. At my last attempt, I decided to disable SRAT in the BIOS. Sure enough, the ESXi 4.0 installation went fine this time. Yay! So the problem could but with ESXi or the perhaps I had a bad BIOS version. (Tyan S3970 ver 2.05) Since this NUMA error is occurring on 2 machines with the same configs, it's hard to tell. IMO, I think it's a ESXi bug. Why? ESXi 3.5 installs and runs fine on the same hardware with SRAT enabled.

[esxtek]

Thanks for the information.

Does this happen with fresh installs of ESXi 4.0 or just with upgrade?

[meetom]

Initially, I did an upgrade from ESXi 3.5 to 4.0i and ran into this error. I eventually build an identical machine for a fresh install of ESXi 4.0 and still had this error. Thus, this happens on both upgrade and fresh install.

[grizzley01]

I suppose you have 2 microprocessors (cpu)

Spread your DIMMs equally over both microprocessor (cpu) DIMM slots (1-8) and (9-16). That means the slots (1-8) should have about the same amount of RAM (GB) as the slots (9-16). This ensures that both microprocessors (cpu) are balanced out for non-uniformed memory access (NUMA). If you have no memory in the second micorprocessor (cpu) DIMM slots (9-16) or a big difference, you get the error: NUMA node 1 has no memory.

[elgordojimenez]

Hello,

It may be too late but the solution is to disable NUMA from the BIOS, see here http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=100369... and check this PDF and search for "NUMA" :

http://www.vmware.com/pdf/Perf_Best_Practices_vSphere4.0.pdf

Cheers.