On Feb 8, 2011, at 17:51 , Pasi Kärkkäinen wrote:
> On Tue, Feb 08, 2011 at 05:28:35PM +0100, Pim van Riezen wrote:
>> On Feb 8, 2011, at 17:10 , Pim van Riezen wrote:
>>
>>>
>>> On Feb 8, 2011, at 16:53 , Pasi Kärkkäinen wrote:
>>>
>>>> Have you gived dom0 fixed amount of memory, and also increase dom0 vcpu
>>>> weights
>>>> so that dom0 will always get enough cpu time to take care of things?
>>>
>>> Fixed dom0_mem, yes.
>>> Weighting, seems not, but just did a testrun with the dom0 weight set to
>>> 512. I got 2 task blocks on one node (that seems a new development) and the
>>> same plus another raw_safe_halt soft lockup on the other:
>>
>> Also tried pinning 2 cpus for domain-0. Still soft lockups.
>>
>
> Did you also make sure VMs don't use those 2 pcpus dedicated for dom0?
> You have to explicitly configure each VM not to use those pcpus.
That seems to have done the trick.
Added to xen command line: dom0_max_vcpus=2 dom0_vcpus_pin
Then tested after running this command:
xm list | ( read && read && cat ) | cut -f1 -d" " | while read guest; do xm
vcpu-pin $guest 0 2-23; done
No soft-lockups. Will do a longer test now. If something new comes up I will
report.
Cheers,
Pim
_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel
|
Home