Windows 2008 R2 VM not restarting correctly after manual Failover

[LuigiProv]

I'm having an issue where my W2K8 R2 VMs are not shutting down / starting up correct after a failover.

I have 2 Servers in my Cluster (2 x HP DL380 G7) running on ESXi 5.0.

The Management Server is a HP DL180 G6 running VCenter 5.0.

I have both VMotion & HA configured.

All the VMs are running from my FC Disk Array (HP P2000) and both my ESXs have access to the LUNs configured on my Disk Array.

So we are testing the HA Failover..

What we do is, we have an x amount of VMs running on ESX1 and an x amount  of VMs running on ESX2.

We manually turn off power on ESX2 to see if the VMs running on ESX2 migrate correctly to ESX1.

The VMs DO migrate to ESX1 after turning off ESX2.

But when they are migrated and restarted on ESX1, that VMs will tell you something on the lines of: "Windows was not shutdown correctly. Do you want to start in safe mode / normal mode / command prompt etc."

Any one has an idea of what might be the cause of this? I have never encountered this before.

I checked with a colleague of mine, and he has not seen this before either.

All the VMs in question have VMWare tools installed on them.

And the settings for VM Monitoring are set for both "VM" and "Application Monitoring".

Thanks in advance....

[kjb007]

The windows machines in an HA event do crash, but they are restarted automatically, albeit the BIOS post and such are skipped.

VM Monitoring instructs the host to watch for the VM  tools heartbeat, and restart the vm in the case of a threshhold being  reached.

What behavior have you not encountered before?

-KjB

[vMario156]

Thats the normal behavior of Windows, if you just power off the system (doesnt matter if virtual or physical).

But as default option it should be selected that Windows starts in normal mode, which gets selected after 30 seconds automatic.

Regards,

Mario

[LuigiProv]

What I hadn't encountered before is the "windows crashing after an HA event".

So based on what vMario156 just said, this is how it "works"?

What about possibility of system corruption (corrupted files etc.) happening because of these "improper shutdowns"?

[a_p_]

The behavior is about the same as if you'd pull the power plug on a physical Windows host. Usually Windows will "survive" but since the memory contents are lost and a lot of files were open at the time the OS crashed you have to expect issues with the OS in some cases.

André

[kjb007 · Accepted solution]

That is correct.  An HA event will restart a vm.  It can not stop it from the crash.  Windows will crash.  The NTFS journal will typically protect from filesystem issues, as it does with physical windows.  HA just gives you hardware fault protection, it gives you high availability, but not fault tolerance, per se.

VMware FT will run a vm on two different hosts at the same time to provide protection against this, but you are limited by vCPU.

-KjB

[LuigiProv]

@kjb007

I guess that answers my question and the question I was going to ask next about FT.

FT would be the best way to go but would be more expensive as you would have to have 2 x the hardware to run the same VMs on 2 ESXs at the same time to provide the same options that Windows Clustering would provide.

I guess Microsoft has some more work to do on their Windows Architecture now that everything is going Virtual.

Thank you all for the quick replies...

I'll just go into windows settings and set it to log on after 5 seconds (as opposed to the normally 30 sec) to have a quicker startup in case of a Fail Over.

[kjb007]

The protection level in order of protection that i see is Vmware HA -> MSCS -> VMware FT

HA protects you from hardware failure, and performs a restart.  MSCS goes a step further since you have two OS's up and running, and will failover resources and start the app. FT maintains two running copies of a VM, go it goes above and beyond MSCS.  What it does not protect against is application failures.  If an app dies, which MSCS will watch for, neither VMware HA nor VMware FT will help.

Hope that helps.

-KjB

[LuigiProv]

Thanks you for explaning that kjb007.

Appreciate it!!!