topic Re: VM start failed in Operating System - HP-UX

VM start failed

N Ward — Thu, 30 Jul 2009 07:26:18 GMT

Hi... A VM Guest terminated overnight and wouldn't start this morning citing not enough memory on the VM Host to start. The Guest is allocated 3Gb from the Host. I was able to start the VM by using hpvmmodify and setting memory to 2Gb.

There were no problems identified on the VM Host, so my question is why would the VM Host use memory set aside for VMs? Is there any way to avoid this in the future?

I haven't tried re-enabling the 1Gb to the VM Guest again yet. Would this require a reboot to become effective since the VM is currently running. Many thanks...

Re: VM start failed

Steven E. Protter — Thu, 30 Jul 2009 08:05:47 GMT

Shalom,

Since your vm is running, you can take a look at a few things.

The logs in /var/adm/syslog

They may contain a clue as to what happened and when.

/var/adm/crash

There may be a crash dump there which can be sent to HP for analysis. If there is a crash dump, HP can determine if a patch is needed on the VM.

SEP

Re: VM start failed

N Ward — Thu, 30 Jul 2009 08:10:15 GMT

Thanks, it has created a dump, but I'm more interested in why it didn't want to start with its allocated memory? I want to understand the relationship between VM Host memory and VM Guest memory, ie Does the VM Host set aside memory that is configured for use with a VM Guest or can it poach it and thus prevent the Guest from starting. This VM Guest has been fine in the past starting with its allocated 3Gb memory but didn't want to today because it said not enough was available on the VM Host.

We are running VM 2.0 so HP are not interested in helping us at the moment until we upgrade to a supported version (which is being planned).

Re: VM start failed

Steven E. Protter — Thu, 30 Jul 2009 08:31:10 GMT

The reason it created the dump and failed to allocate the memory could be linked.

Normally performing q4 analysis on the crash dump leads to installation of a patch recommended by HP software support. In this case, based on the memory problems it could b a critical, memory related patch.

This thread contains two ways of doing the q4 analysis.
http://forums11.itrc.hp.com/service/forums/questionanswer.do?threadId=617986

Don't ignore the evidence you have, the memory problem and the crash could be very closely related.

:-)

SEP

Re: VM start failed

N Ward — Thu, 30 Jul 2009 09:16:43 GMT

Thanks that may well be an option. Do you know the relationship though between VM Host memory and VM Guest memory. Can a VM host use memory for itself that is/would be allocated to a VM Guest had it been up?

Re: VM start failed

Steven E. Protter — Thu, 30 Jul 2009 09:28:28 GMT

Shalom again,

Can a VM host use memory for itself that is/would be allocated to a VM Guest had it been up?

My understanding is that once allocated to the guest, the vm host is not going to use the memory for any other purpose.

To try and do so would probably lead to a memory fault and crash.

So its not supposed to happen. But then again this is exactly what you experienced.

I am thinking there may be a patching issue for the VM host.

What version of VM is in use? What is the Host OS? What QPK is installed?

SEP

Re: VM start failed

N Ward — Thu, 30 Jul 2009 09:34:47 GMT

Yes it would seem illogical to me too.

We are using VM Integrity version 2.0. The host is Itanium running 11.23.0609

We are going through an exercise at the moment in preparation for an update to the VM Software to 3.5. The pre-req for this is that the VM Host OS be at least 11.23.0712. I've already ran the SWA analysis and prepared the install, but cannot go ahead until tested on some DEV kit first.

Re: VM start failed

Steven E. Protter — Thu, 30 Jul 2009 09:53:34 GMT

Shalom again,

I recommend a process that gets you to the most current version of VM possible. I know 4.1 is released, but have not looked up the OS requirements.

HP has made great strides in VM with recent releases and you may find more stability and functionality by continuing down your current path.

I would definitely do 14 analysis and have the output analysed by HP. It may illuminate a number of issues.

SEP

Re: VM start failed

Don Morris_1 — Thu, 30 Jul 2009 10:03:22 GMT

A Guest is just a process (with some special hooks) in the Host. The memory assigned to it is locked (like mlock for other processes) so the Host doesn't modify it.

Once the Guest terminated, however -- the memory is released to the Host... so yes, _then_ the memory could be used for other purposes (like buffer/file cache most likely if the Host isn't being used to do other work as recommended).

Re: VM start failed

N Ward — Thu, 30 Jul 2009 10:29:51 GMT

So if it were in use and a VM Guest start was requested (not a new VM, an existing one) it could deny access to the memory?