Hi all. Sometimes different HVM guests (but all of them have vcpus>1) stuck with no state flags #xm vcpu-list w157 Name ID VCPU CPU State Time(s) CPU Affinity w157 507 0 6 --- 64864.1 0-11 w157 507 1 0 -b- 34287.3 0-11 or #xm vcpu-list w221 Name ID VCPU CPU State Time(s) CPU Affinity w221 571 0 18 --- 26898.9 12-23 w221 571 1 20 --- 27045.7 12-23 Maybe this is some kind of scheduler bug when running a lot of (50+) guests, e.g. depends on amount of qemu-dm processes on dom0? dom0: debian wheezy x86_64/xen 4.1.3, CPU 2xE5-2620 domU: Windows 2008R2 dom0 weight is 1024, all guests - 256 with different cap. _______________________________________________ Xen-users mailing list Xen-users@lists.xen.org http://lists.xen.org/xen-users
On 01/13/2013 08:34 AM, Igor Pavlikevich wrote:> Hi all. > > Sometimes different HVM guests (but all of them have vcpus>1) stuck > with no state flags > #xm vcpu-list w157 > Name ID VCPU CPU State Time(s) CPU Affinity > w157 507 0 6 --- > 64864.1 0-11 > w157 507 1 0 -b- > 34287.3 0-11 > or > #xm vcpu-list w221 > Name ID VCPU CPU State Time(s) CPU Affinity > w221 571 0 18 --- > 26898.9 12-23 > w221 571 1 20 --- > 27045.7 12-23 > > > Maybe this is some kind of scheduler bug when running a lot of (50+) > guests, e.g. depends on amount of qemu-dm processes on > dom0? > dom0: debian wheezy x86_64/xen 4.1.3, CPU 2xE5-2620 > domU: Windows 2008R2 > dom0 weight is 1024, all guests - 256 with different cap.Hi Igor, these domUs without ''state'' shown are ok. That''s because the xm command is reading the state of the domU from xend/XenHypervisor and once it will not get any answer from domU the state will just not display. This means that your domUs or dom0 are highly loaded and didn''t provide the state output for the limited amount of time. From the xm man page: <cite> Most xm commands act asynchronously, so just because the xm command returned doesn''t mean the action is complete. This is important, as many operations on domains, like create and shutdown, can take considerable time (30 seconds or more) to bring the machine into a fully compliant state. If you want to know when one of these actions has finished you must poll through xm list periodically. </cite> And this is valid also for most ''xm *list'' commands. Best regards, -- Peter Viskup
Hi all. On Sun, Jan 20, 2013 at 10:44 PM, Peter Viskup <skupko.sk@gmail.com> wrote:> On 01/13/2013 08:34 AM, Igor Pavlikevich wrote: > >> Hi all. >> >> Sometimes different HVM guests (but all of them have vcpus>1) stuck with >> no state flags >> #xm vcpu-list w157 >> Name ID VCPU CPU State Time(s) CPU Affinity >> w157 507 0 6 --- 64864.1 >> 0-11 >> w157 507 1 0 -b- 34287.3 >> 0-11 >> or >> #xm vcpu-list w221 >> Name ID VCPU CPU State Time(s) CPU Affinity >> w221 571 0 18 --- 26898.9 >> 12-23 >> w221 571 1 20 --- 27045.7 >> 12-23 >> >> >> Maybe this is some kind of scheduler bug when running a lot of (50+) >> guests, e.g. depends on amount of qemu-dm processes on >> dom0? >> dom0: debian wheezy x86_64/xen 4.1.3, CPU 2xE5-2620 >> domU: Windows 2008R2 >> dom0 weight is 1024, all guests - 256 with different cap. >> > > Hi Igor, > these domUs without ''state'' shown are ok. That''s because the xm command is > reading the state of the domU from xend/XenHypervisor and once it will not > get any answer from domU the state will just not display. This means that > your domUs or dom0 are highly loaded and didn''t provide the state output > for the limited amount of time. > > From the xm man page: > <cite> > Most xm commands act asynchronously, so just because the xm command > returned doesn''t mean the action is complete. This is important, as many > operations on domains, like create and shutdown, can take considerable time > (30 seconds or more) to bring the machine into a fully compliant state. If > you want to know when one of these actions has finished you must poll > through xm list periodically. > </cite> > > And this is valid also for most ''xm *list'' commands. > >I understand this but in my case vcpu with no state never returns to block/running/etc state and usage timer of this vcpu does not increase, so HVM completely stuck. _______________________________________________ Xen-users mailing list Xen-users@lists.xen.org http://lists.xen.org/xen-users