[test day] Windows SBS 2011 random hang and resource utiluization on 4.4-rc2 and rc3

[test day] Windows SBS 2011 random hang and resource utiluization on 4.4-rc2 and rc3

[Don Brearley]

Hello, 


I know it's not officially "test day" but I've been toying with Xen 4.4-rc2 and -rc3 and I have several successful linux
PV guests running on 3.13.1.  Works fantastic.


However, I have an SBS 2011 64bit HVM installation (no GPLPV drivers) and I routinely get a complete hang.  The VM 
goes 'stateless' in 'xl list' (shows the domain state as ------).  Cannot access via RDP, cannot access via VNC, cannot
ping, etc.   Just totally dead, yet the VM "lives".  Issing 'xl shutdown -F domain' does not kill it, I have to use
"destroy" to remove the domain.  I didnt install GPLPV because I didnt see they were supported for SBS 2011.


No errors in the xen or qemu logs, nothing in the event log, nothing in dmesg/messages on dom0.


I can't seem to find anything to trigger it... it just, happens, seemingly at random. 


"xl top" shows the domain as using 200% CPU when this happens.... in one case I left the VM running like that
overnight, and returned, and it had come up to 300% utilization.


The domain sits on a physical disk (/dev/sdd) which is replicated via DRBD to another server. 


Any hints? Suggestions?




The domU config:  (typed it, didnt copy/paste).


builder="hvm"
memory="32767"
maxmem="32767"
shadow = "128"
vcpu = 4
viridian = 1


vif = [ 'model=e1000, bridge=xen-pribr' ]


disk = [ 'phy:/dev/sdd,xvda,w' ]


videoram=128

Re: [test day] Windows SBS 2011 random hang and resource utiluization on 4.4-rc2 and rc3

[Andrew Cooper]

On 06/02/14 18:03, Don Brearley wrote:
> Hello, 
>
>
> I know it's not officially "test day" but I've been toying with Xen 4.4-rc2 and -rc3 and I have several successful linux
> PV guests running on 3.13.1.  Works fantastic.
>
>
> However, I have an SBS 2011 64bit HVM installation (no GPLPV drivers) and I routinely get a complete hang.  The VM 
> goes 'stateless' in 'xl list' (shows the domain state as ------).  Cannot access via RDP, cannot access via VNC, cannot
> ping, etc.   Just totally dead, yet the VM "lives".  Issing 'xl shutdown -F domain' does not kill it, I have to use
> "destroy" to remove the domain.  I didnt install GPLPV because I didnt see they were supported for SBS 2011.
>
>
> No errors in the xen or qemu logs, nothing in the event log, nothing in dmesg/messages on dom0.
>
>
> I can't seem to find anything to trigger it... it just, happens, seemingly at random. 
>
>
> "xl top" shows the domain as using 200% CPU when this happens.... in one case I left the VM running like that
> overnight, and returned, and it had come up to 300% utilization.
>
>
> The domain sits on a physical disk (/dev/sdd) which is replicated via DRBD to another server. 
>
>
> Any hints? Suggestions?

Boot with "guest_loglvl=all" on the Xen command line.

`xl debug-keys q` and `xl dmesg` to see what is reported from Xen as far
as the guest is concerned

`xen-hvmctx $domid` to see the architectural state of it.

Both be useful as a starting point.

~Andrew

Re: [test day] Windows SBS 2011 random hang and resource utiluization on 4.4-rc2 and rc3

[Don Brearley]

>>> Andrew Cooper  02/06/14 12:09 PM >>>
On 06/02/14 18:03, Don Brearley wrote:
> Hello, 
>
>
> I know it's not officially "test day" but I've been toying with Xen 4.4-rc2 and -rc3 and I have several successful linux
> PV guests running on 3.13.1.  Works fantastic.
>
>
> However, I have an SBS 2011 64bit HVM installation (no GPLPV drivers) and I routinely get a complete hang.  The VM 
> goes 'stateless' in 'xl list' (shows the domain state as ------).  Cannot access via RDP, cannot access via VNC, cannot
> ping, etc.   Just totally dead, yet the VM "lives".  Issing 'xl shutdown -F domain' does not kill it, I have to use
> "destroy" to remove the domain.  I didnt install GPLPV because I didnt see they were supported for SBS 2011.
>
>
> No errors in the xen or qemu logs, nothing in the event log, nothing in dmesg/messages on dom0.
>
>
> I can't seem to find anything to trigger it... it just, happens, seemingly at random. 
>
>
> "xl top" shows the domain as using 200% CPU when this happens.... in one case I left the VM running like that
> overnight, and returned, and it had come up to 300% utilization.
>
>
> The domain sits on a physical disk (/dev/sdd) which is replicated via DRBD to another server. 
>
>
> Any hints? Suggestions?

Boot with "guest_loglvl=all" on the Xen command line.

`xl debug-keys q` and `xl dmesg` to see what is reported from Xen as far
as the guest is concerned

`xen-hvmctx $domid` to see the architectural state of it.

Both be useful as a starting point.

~Andrew

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel




Hi Andrew,


I do boot xen with "multiboot /boot/xen-4.gz dom0_mem=4096M loglvl=all guest_loglvl=all"


I will try those commands you suggested when I enounter the next hang, and will report it
back to you and the list.


Thanks much!
- Don