Bios settings?

[s00pa]

Hi,
without going into huge amount of details. I have a windows 2008 guest which when run on an IBM server runs perfectly fine (rock-solid, no blue screens) but running on Intel server has blue screens every couple of days (main 0x3B errors). I have already done a lot of googling for the errors to see how to stabilise the server, without any success.
Before going into any details on the configuration, the main question I have are what bios settings could potentially be causing this issue?

Comparing the two servers' bios's is not easy as some settings are not on both and there are extra settings, etc. but what should the following be set to?

Enhanced Intel Speedstep Tech (currently enabled on Intel server)
Processor C3 (disabled on intel server)
Processor C6 (enabled on intel server)
Execute Disable Bit (enabled on intel server / disabled on IBM server)
Intel VT for directed I/O (disabled on Intel Server)

Other misc settings or anything I've missed (apart from virtualization setting ) ?? (MLC streamer, MLC spatial prefetcher, DCA)

I was going to try a few things out but thought that getting proper advice first would be better

Thanks in advance,

Sean

 

[spirit]

Hi,
without going into huge amount of details. I have a windows 2008 guest which when run on an IBM server runs perfectly fine (rock-solid, no blue screens) but running on Intel server has blue screens every couple of days (main 0x3B errors). I have already done a lot of googling for the errors to see how to stabilise the server, without any success.
Before going into any details on the configuration, the main question I have are what bios settings could potentially be causing this issue?

Comparing the two servers' bios's is not easy as some settings are not on both and there are extra settings, etc. but what should the following be set to?

Enhanced Intel Speedstep Tech (currently enabled on Intel server)
Processor C3 (disabled on intel server)
Processor C6 (enabled on intel server)
Execute Disable Bit (enabled on intel server / disabled on IBM server)
Intel VT for directed I/O (disabled on Intel Server)

Other misc settings or anything I've missed (apart from virtualization setting ) ?? (MLC streamer, MLC spatial prefetcher, DCA)

I was going to try a few things out but thought that getting proper advice first would be better

Thanks in advance,

Sean

Enhanced Intel Speedstep Tech : off

This cause cause some clock speed problem and blue screen on windows.


(Try to disable options which can change speed, core disabling ,etc... )

 

[s00pa]

Thanks. I appreciate the reply .. I'm going to try changing a few settings tonight and see how it goes.

 

[s00pa]

I tried changing the bios settings but I don't think it made a difference. In fact I'm finding a pattern which is troublesome if true. I was finding Windows Server 2008 was blue screening (mainly 0x3b) every day or two. After changing the bios settings and restarting the entire server, it was stable for about 10 days and then 0x3b every couple of days again. I thought I'd try rebooting the entire proxmox server again and once again the windows guest was rocksolid for around another 10 days and then back to 0x3b 2 days in a row. I have rebooted the server again this morning but I'm stumped as to why this would be occurring.

--Sean

 

[spirit]

I tried changing the bios settings but I don't think it made a difference. In fact I'm finding a pattern which is troublesome if true. I was finding Windows Server 2008 was blue screening (mainly 0x3b) every day or two. After changing the bios settings and restarting the entire server, it was stable for about 10 days and then 0x3b every couple of days again. I thought I'd try rebooting the entire proxmox server again and once again the windows guest was rocksolid for around another 10 days and then back to 0x3b 2 days in a row. I have rebooted the server again this morning but I'm stumped as to why this would be occurring.

--Sean

mmm, this is strange. Maybe could it be a hardware memory problem ? (do you have tried to do a memtest ?)

 

[s00pa]

mmm, this is strange. Maybe could it be a hardware memory problem ? (do you have tried to do a memtest ?)

Thanks for the reply.. I certainly thought it quite odd too. I ran memtest for close on 24 hours with no problems
On a side-note. if this pattern does seem to be consistent, it's not a perfect solution by any means but is it easy to schedule proxmox to shutdown virtuals and do a full server restart once a week? Can I set as cron or is there a better way?

--
Sean

 

[snowman66]

Does this only affect windows machines?
Check Event Viewer -> System/Application/Security to see what is happening.

There are some hotfixes:
http://support.microsoft.com/kb/2584454
http://support.microsoft.com/kb/2359223
http://support.microsoft.com/kb/2639032/en-us
http://search.microsoft.com/en-us/supportresults.aspx?form=mssupport&q=0x0000003B

Processor C3 (disabled on intel server)
Processor C6 (enabled on intel server)

This are power saving modes.

 

[spirit]

Does this only affect windows machines?
Check Event Viewer -> System/Application/Security to see what is happening.

There are some hotfixes:
http://support.microsoft.com/kb/2584454
http://support.microsoft.com/kb/2359223
http://support.microsoft.com/kb/2639032/en-us
http://search.microsoft.com/en-us/supportresults.aspx?form=mssupport&q=0x0000003B

Processor C3 (disabled on intel server)
Processor C6 (enabled on intel server)

This are power saving modes.

Always disable power saving features, dynamic frequency,etc....

The problem is that if a core is shutted down, or frequency change, the vms are unware of this, and this can lead to problems, mainly clock problems.
And windows guests really don't like this.

(Note: Maybe it can be improved soon with qemu hyper-v features, but i'm not sure)

 

[s00pa]

Does this only affect windows machines?
Check Event Viewer -> System/Application/Security to see what is happening.

There are some hotfixes:
http://support.microsoft.com/kb/2584454
http://support.microsoft.com/kb/2359223
http://support.microsoft.com/kb/2639032/en-us
http://search.microsoft.com/en-us/supportresults.aspx?form=mssupport&q=0x0000003B

Processor C3 (disabled on intel server)
Processor C6 (enabled on intel server)

This are power saving modes.

Thanks, I really appreciate this information. Unfortunately I'm not running 2008 _R2_ so most of these fixes don't apply afaik. The part that confused the hell out of me is that the virtual was perfectly stable on another proxmox server (an IBM server). I have adjusted all the bios settings that could alter the clock but still have the instability.. The only thing I can put my finger on is that it appears to occur when ShadowProtect does its backup (I have it set for images every 30 minutes) .. on another server I have managed to keep it _more_ stable but decreasing the number of images to a single image per day but it still blue screens every now and then, just not as often.
Just as extra info, I have tried virtio vs legacy drivers for HDD and network with no difference. also pveversion (pve-manager: 2.2-32 (pve-manager/2.2/3089a616)) so I know I should probably update to latest version before posting here but was just hoping to find a "me too" with solution

--
Sean

 

[spirit]

Thanks, I really appreciate this information. Unfortunately I'm not running 2008 _R2_ so most of these fixes don't apply afaik. The part that confused the hell out of me is that the virtual was perfectly stable on another proxmox server (an IBM server). I have adjusted all the bios settings that could alter the clock but still have the instability.. The only thing I can put my finger on is that it appears to occur when ShadowProtect does its backup (I have it set for images every 30 minutes) .. on another server I have managed to keep it _more_ stable but decreasing the number of images to a single image per day but it still blue screens every now and then, just not as often.
Just as extra info, I have tried virtio vs legacy drivers for HDD and network with no difference. also pveversion (pve-manager: 2.2-32 (pve-manager/2.2/3089a616)) so I know I should probably update to latest version before posting here but was just hoping to find a "me too" with solution

--
Sean

Can you try to install last proxmox kernel from pve-test repository, and add to your vm config file:

args: -cpu kvm64,hv_relaxed

(and remove cpu: if you have one)


this will add hyper-v feature to tell to windows that it's a virtual machine, and should help to avoid bsod on heavy load.