stack/hopr

Clearing obstacles of RAID, kernel panic and hostname to get the ESX Server serving

It was a normal day at office. Slightly bored with work at hand, I got interested in a problem my co-wrokers were facing. It was a Dell Precision 670. They installed ESX Server 3.5 on it. But, it won’t get past GRUB.

I walked into the lab. GRUB was stuck like this:
GRUB _

It was no prompt. Rather, it looked like grub was stuck at something. And in fact, it was stuck. It was seeing red in RAID. There were two 80GB hard drives in there configured at RAID0. I disabled RAID and the booting got past GRUB.

But, the troubles didn’t end there. The kernel panicked.

WARNING: VMK: 538: Initialization of vmkernel failed, status 0xbad0001e

After a reinstall, few reboots and a lot of Google hits, I saw some light.

CPUID is an instruction to determine the CPU type and its features. Intel processors from the Pentium Pro onwards have a maximum CPUID input value of only 02h or 03h. The only exception is the Intel Pentium 4 with Hyper-Threading Technology (HTT). Older operating systems released before the Intel Pentium 4 with HTT and are therefore not aware of such a processor. They are not capable of handling the extra CPUID information provided by the processor with EAX input values greater than 03h. The ESX Server kernel seems as such. And the Dell Precision 670 has Intel HT processor(s). So, we add the following line at the end of file: /etc/vmware/esx.conf:

/vmkernel/checkCPUIDLimit = “false”

One mistake I repeatedly did was that I used the same console as used to boot. That is, as soon as the vmkernel threw the error and gave me a prompt, I used the same prompt to do the changes. Only to be perplexed later that the kernel threw the same error. And it lost the changes made to esx.conf.

Looking at the boot options:

VMWare ESX Server
VMWare ESX Server (debug mode)
Service Console only (troubleshooting mode)

It dawned on me that VMWare has provided with a service console with some purpose. I used this console to make the changes and rebooted. Success. VMWare booted successfully and displayed the welcome screen.

Now, back to the desk. Connecting to ESX Server remotely and uploading the virtual images. While uploading the images through VMWare Infrastructure Client, there was another error:

“the host name used for the connection does not match the subject name on the host certificate”

I had named the server as “esx” when I installed it. And the /etc/hosts file looked like this:

127.0.0.1 localhost.localdomain localhost
10.217.92.60 esx

Adding another entry, telling the server it’s own name does the job. We’re just letting the ESX server resolve it’s own name.

127.0.0.1 localhost.localdomain localhost
10.217.92.60 esx
127.0.0.1 esx

Finally, ESX Server is serving up the operating systems. Yummy!