Recursos para Profesionales de TI >
Página principal de foros
>
Hyper-V
>
CPU in VM will at random times max out for 20-30min then go back to normal
CPU in VM will at random times max out for 20-30min then go back to normal
- I have a VM that runs good most of the time, but at other seemingly random times the CPU will suddenly spike to 100% for about 20-30min before returning to normal.
The task manager in the VM does not show anything unusual, so I'm thinking the VM is suddenly getting less resources. only problem is that this VM is the only VM on the entire box. The node itself of course shows little resource usage. I also notice in performance manager C3 idle time in performance monitor is near 80% when this is occuring, yet the VM is maxed out. VMM shows the VM is using ~25% when the VM itself is reporting 100%. The thing is this VM is not using a lot of resources, I am sure of that...
any ideas???
thank you
Respuestas
- Here are the basics on checking and comparing the revisions:
http://itproctology.blogspot.com/2008/04/determining-versions-under-hyper-v.html
Did you try dsiabling TCPOffloading?
Have you attempted changing any of the settings of the VMs to see if that has any effect (single processor, 1024 MB RAM)?
Did you mention if there were any apps running in the parent partition? You log out of the parent and RDP direct into the VM and the behavior stays the same?
Personally, I am leaning toward a combination of the application that is running and the configuration of the VM.
Brian Ehlert (hopefully you have found this useful)- Marcado como respuestaBarrySDCA martes, 13 de octubre de 2009 6:53
Todas las respuestas
- First - I think it is important to understand the Vm itself.
What applications are running within the VM?
antivirus?, processing applications? workflow applications? etc.
Is the VM literlly idle? No installed applications at all, only the base operating system install??
what is the OS of the VM?
Then we move to the host:
Is there any disk I/O happening during this perceived pause?
Is the host doing anything? Possibly AV running on the host that is demanding resources and thus starving the VM?
Brian Ehlert (hopefully you have found this useful) - Hi Brian,
I appreciate your reply to help me. This is a frustrating issue and I have been going in circles on it....
I have a bunch of appliaitons running in the VM. Each app does use some disk to record logs and minor but constant IP activity to receive and process info. The CPU would typically be doing 30-40% on a standalone box, and there may be brief periods of some bursting but nothing near 100%.
OS is 2003 R2 x64 enterprise
this is the only VM on the node. I set it up this way so I could live migrate them instead of rebooting the apps, which is a real PITA.
the host is doing absolutely nothing but run the VM. a/v scan is not set to go until saturday, so that can't be it (A/V i nthe VM).
THANK YOU - more info....
I found in the event log of the 2003 server that is running in the VM:
The miniport 'Microsoft Virtual Machine Bus Network Adapter' hung
Why would this be hanging?
thank you - Have you verified that the revisions of the Integration Services installed in the VM and running on the host match?
Brian Ehlert (hopefully you have found this useful) - I installed integration services from the same host the VM is on now and have not migrated it anywhere, so I assume it is. How can I verify?
I'm actually experiencing this problem on every host I have it installed in
the host is 2008R2
the VM is 2003 R2
what do you think?
thank you - Here are the basics on checking and comparing the revisions:
http://itproctology.blogspot.com/2008/04/determining-versions-under-hyper-v.html
Did you try dsiabling TCPOffloading?
Have you attempted changing any of the settings of the VMs to see if that has any effect (single processor, 1024 MB RAM)?
Did you mention if there were any apps running in the parent partition? You log out of the parent and RDP direct into the VM and the behavior stays the same?
Personally, I am leaning toward a combination of the application that is running and the configuration of the VM.
Brian Ehlert (hopefully you have found this useful)- Marcado como respuestaBarrySDCA martes, 13 de octubre de 2009 6:53
- Hello,
I'm having the same problem on all my Hyper-V machines.
The miniport 'Microsoft Virtual Machine Bus Network Adapter' hung etc etc
I have 6 hosts an Dell PE2950.
Did you find any solution?
Thanks
LinkState - Has a solution been found for this?
I literally am having the same issue.
I too am getting this error when the server is maxed out: The miniport 'Microsoft Virtual Machine Bus Network Adapter' hung
I've got a Windows Hyper-V Server 2008 R2 host running a Windows Server 2003 SP2 DC and also a CentOS guest. The linux vm seems to work perfectly when this occurs and is unaffected but the windows guest will crank on the CPU at 100% for around 20-30 mins.
I have also noticed that if I simply pause the windows vm and then resume, it corrects the problem. It also fixes it if I reboot the Host server. I can also confirm that in the MMC console it reports the CPU at only 20% even though the guest is reporting 100% CPU usage.
While the problem is occurring I am able to see in the Task Manger that it's the 'System' process that is taking the majority of CPU % and it always shows 256K memory usage for that process.
I have verified that the correct version of Integrated Services is running on the guest and have matched up with the version of vmms.exe on the host
I am not really sure what is meant by disabling TCPOffloading. I would like to try that if it might help but do I disable it on the host or the guest? Some information of that would be helpful for troubleshooting this issue. - TCPOffloading has been discussed quite a bit in this forum.
You disable the setting in the driver of the Host NIC that is attached to the External Virtual Network.
And, there are many TCPOffloading options - this is not the same as disabling TCP Chimey, which is rarely necessary.
Brian Ehlert (hopefully you have found this useful) - Okay I'm looking into disabling TCPOffloading now.
First I disabled the virtual network adapter in the primary partition since it is unused and is getting a DHCP address. Could this be the cause of the problem?
I'm searching the net to figure out how to disable TCPOffloading on a Server Core install or the free Hyper-V 2008 R2 server (Which is command line only) but I haven't came up with anything yet.
From what I can tell TCPOffloading is disabled within the device properties of your physical NIC normally found under the device manager but how is this possible with no GUI?
Thanks in advance. - Well disabling the virtual nic did nothing, problem still exists. The only way I can tell to disable TCPOffloading in Hyper-v Server and Server Core is to use the registry editor.
Using a program called DeviceRemover I can get to the exact location in the Registry for my physical nic on the primary partition. I just don't know which options to disable. I have several options with the word Offloading so I will list them:
IPChecksumOffloadIPv4
LsoV2IPv4 (Large Send Offload v2 (IPv4))
LsoV2IPv4 (Large Send Offload v2 (IPv6))
TCPChecksumOffloadIPv4
TCPChecksumOffloadIPv6
UDPChecksumOffloadIPv4
UDPChecksumOffloadIPv6
Should I disable all of them? I am going to try using a legacy nic to help troubleshoot the issue while I am waiting on a response. disable all that you don't need.
If you don't use IPv6 - then just disable to protocol all together.
Brian Ehlert (hopefully you have found this useful)- Hi,
It seems that the problem was coming from the Hyper-V integration services installed on guest machines.
On the Host machines that were migrated from Win2008 to Win2008 R2, the guest machines have 2 versions of integration services installed, one it's called Hyper-V Integration Services (version blah blah) and the other it's called Hyper-V Guest Components. Somehow when you try to upgrade the addons, it installs a 2nd version instead of upgrading the one already installed.
I removed both of them, rebooted and reinstalled the last version and there's no hang anymore.
Regards,
LinkState- Propuesto como respuestaLinkState viernes, 13 de noviembre de 2009 8:45
- Thanks for the response LinkState. Unfortunately for me this server was not a migrated guest but it was a p2v conversion... I only have a single instance of Hyper-V Integration Services (version 6.1.7600.16385) listed under Add/Remove programs however I do feel like this may somehow be related to the Hyper-V Integration Services version or that it may need an update or is somehow driver related since this is a p2v vm.
The Hyper-V host server is the 2008 R2 (64-bit) version and I've never ran this vm on any other server. I have also verified that the version of integration services on the host and the guest match by following this TechNet Page: http://technet.microsoft.com/en-us/library/ee207413(WS.10).aspx
I have now disabled the TCPOffloading on the host as discussed and disabled all IPv6 as well but I'm not too optimistic.
I think I will try uninstalling and reinstalling the guest Integration Services if it's still not fixed after this.
I will continue to post updates until this is resolved. - So it has been 10 days since I disabled TCPOffloading & IPv6 in the registry on the host (primary partition) and it hasn't done it since...
Could the problem really be fixed!? I will post back again if something happens or if it's still problem free after a while.
Thanks for everyone's help on this issue! - Interesting. I would not expect it to be related at all. But your experience indicates otherwise.
Be sure to keep watching to make sure that there was not just an application use pattern change.
I will forward this thread on to the team to see if they have any comments.
Thanks!
Brian Ehlert (hopefully you have found this useful) acjohnson:
Could you please fill us in on the brand and revision of the NIC in your server?
I am wondering if this relates to a driver or a specific NIC.
Brian Ehlert (hopefully you have found this useful)- In my server I am running a Tyan Tempest i5400XT (S5396) motherboard (which is not technically supported yet under 2008 R2) and I'm using one of the integrated Intel 82563EB Gigabit NICs for connectivity.
I have not attempted to install NIC drivers straight from Intel yet. I am using the signed driver that came with Hyper-V Server 2008 R2.
I am also considering a BIOS update as this motherboard is still running the stock bios but this has been the only issue so far.
- Well..shoot. You blew that theory out of the water.
But, a BIOS update could cause things to behave better- that is always true.
Brian Ehlert (hopefully you have found this useful)
