[ovirt-users] Problem starting VMs

Wolfgang Bucher wolfgang.bucher at netland-mn.de
Tue Aug 23 14:13:18 EDT 2016


Hello



i just changed xfs to ext4 and defragmantion is much better then before, i will also try kernel lts from elrepo and test it again.



Thanks



Wolfgang



-----Ursprüngliche Nachricht-----
Von: Yaniv Kaul <ykaul at redhat.com>
Gesendet: Die 23 August 2016 19:46
An: Wolfgang Bucher <wolfgang.bucher at netland-mn.de>
CC: Charles Gruener <cgruener at gruener.us>
Betreff: Re: AW: [ovirt-users] Problem starting VMs


On Aug 23, 2016 7:33 PM, "Wolfgang Bucher" <wolfgang.bucher at netland-mn.de <mailto:wolfgang.bucher at netland-mn.de> > wrote:
 >
 > Hello
 >
 >
 > I am using local storage with adaptec raid controller, disk format is raw

Raw is always raw-sparse, so that may explain this somehow, yet still odd that Windows installation would cause so much fragmentation.
 I wonder if using the discard hook (and IDE or virtio-scsi) would help - or perhaps using a qcow2 makes more sense (create a snapshot right after disk creation for example).

>
 >
 > image: e4d797d1-5719-48d0-891e-a36cd4a79c33
 > file format: raw
 > virtual size: 50G (53687091200 bytes)
 > disk size: 8.5G
 >
 >
 > this is a fresh installation of W2012 after installation i got with xfs_db -c frag -r  /dev/sdb1:
 >
 > aktuell 407974, ideal 35, Unterteilungsfaktor 99,99%

And the XFS formatted with default parameters?
 Y.

>
 >
 > Thanks
 >
 >
 > Wolfgang
 >
 >
 >> -----Ursprüngliche Nachricht-----
 >> Von: Yaniv Kaul <ykaul at redhat.com <mailto:ykaul at redhat.com> >
 >> Gesendet: Die 23 August 2016 17:56
 >> An: Wolfgang Bucher <wolfgang.bucher at netland-mn.de <mailto:wolfgang.bucher at netland-mn.de> >
 >> CC: Michal Skrivanek <michal.skrivanek at redhat.com <mailto:michal.skrivanek at redhat.com> >; users at ovirt.org <mailto:users at ovirt.org> (users at ovirt.org <mailto:users at ovirt.org> ) <users at ovirt.org <mailto:users at ovirt.org> >
 >>
 >> Betreff: Re: [ovirt-users] Problem starting VMs
 >>
 >>
 >>
 >> On Tue, Aug 23, 2016 at 6:40 PM, Wolfgang Bucher <wolfgang.bucher at netland-mn.de <mailto:wolfgang.bucher at netland-mn.de> > wrote:
 >>>
 >>> Hello
 >>>
 >>>
 >>> in var log messages i get following :
 >>>
 >>>
 >>> kernel: XFS: possible memory allocation deadlock in kmem_alloc (mode:0x250)
 >>>
 >>>
 >>> I have this problem on 4 different hosts.
 >>>
 >>>
 >>> This happens during copying files from network to a thin provisioned disk, no problems with preallocated disks.
 >>
 >>
 >> What kind of storage are you using? local storage? Even though, it makes little sense to me - the disk is a qcow2 disk, which shouldn't be very fragmented as you might think (qcow2 grows in 64K chunks).
 >> It may grow and grow and grow (until you sparsify it), but that's not going to cause fragmentation. What causes it to be fragmented? Perhaps the internal qcow2 mapping is quite fragmented?
 >> Y.
 >>  
 >>>
 >>>
 >>> Thanks
 >>>
 >>> Wolfgang
 >>>
 >>>
 >>>> -----Ursprüngliche Nachricht-----
 >>>> Von: Michal Skrivanek <michal.skrivanek at redhat.com <mailto:michal.skrivanek at redhat.com> >
 >>>> Gesendet: Die 23 August 2016 17:11
 >>>> An: Wolfgang Bucher <wolfgang.bucher at netland-mn.de <mailto:wolfgang.bucher at netland-mn.de> >
 >>>> CC: Milan Zamazal <mzamazal at redhat.com <mailto:mzamazal at redhat.com> >; users at ovirt.org <mailto:users at ovirt.org> (users at ovirt.org <mailto:users at ovirt.org> ) <users at ovirt.org <mailto:users at ovirt.org> >
 >>>> Betreff: Re: [ovirt-users] Problem starting VMs
 >>>>
 >>>>
 >>>>> On 23 Aug 2016, at 11:06, Wolfgang Bucher <wolfgang.bucher at netland-mn.de <mailto:wolfgang.bucher at netland-mn.de> > wrote:
 >>>>>
 >>>>> Thank's
 >>>>>
 >>>>> but what do you mean with "initialization is finished”
 >>>>
 >>>>
 >>>> until it gets from WaitForLaunch to PoweringUp state, effectively until the qemu process properly starts up
 >>>>
 >>>>>
 >>>>> sometimes the vm crashes while copying files!
 >>>>
 >>>>
 >>>> when exactly? Can you describe exactly what you are doing and what is reported as a reason for crash. When exactly does it crash and how?
 >>>>
 >>>> Thanks,
 >>>> michal
 >>>>
 >>>>>
 >>>>>
 >>>>>
 >>>>> Wolfgang
 >>>>>
 >>>>>
 >>>>>> -----Ursprüngliche Nachricht-----
 >>>>>> Von: Milan Zamazal <mzamazal at redhat.com <mailto:mzamazal at redhat.com> >
 >>>>>> Gesendet: Die 23 August 2016 16:59
 >>>>>> An: Wolfgang Bucher <wolfgang.bucher at netland-mn.de <mailto:wolfgang.bucher at netland-mn.de> >
 >>>>>> CC: users at ovirt.org <mailto:users at ovirt.org> (users at ovirt.org <mailto:users at ovirt.org> ) <users at ovirt.org <mailto:users at ovirt.org> >
 >>>>>> Betreff: Re: AW: Problem starting VMs
 >>>>>>
 >>>>>> Wolfgang Bucher <wolfgang.bucher at netland-mn.de <mailto:wolfgang.bucher at netland-mn.de> > writes:
 >>>>>>
 >>>>>> > the problem is "waiting for launch" takes up to 20 min.
 >>>>>> >
 >>>>>> >
 >>>>>> > I did a lot of tests with some vm's and th problem is:
 >>>>>> >
 >>>>>> >
 >>>>>> > 1.  create a win2012 server
 >>>>>> >
 >>>>>> > 2. attache a new disk (thin provision)
 >>>>>> >
 >>>>>> > 3. fill the disk from network share with about 100G
 >>>>>> >
 >>>>>> > 4. shutdown the vm
 >>>>>> >
 >>>>>> > 5. reboot host and try starting the vm (takes a lot of time until launch)
 >>>>>> >
 >>>>>> >
 >>>>>> >
 >>>>>> > It's because of a high fragmented filessystem 
 >>>>>>
 >>>>>> I see, thank you for the explanation. In such a case, it may take a lot
 >>>>>> of time before all the initialization is finished.
 >>>>>>
 >>>>>> Regards,
 >>>>>> Milan
 >>>>>
 >>>>> _______________________________________________
 >>>>> Users mailing list
 >>>>> Users at ovirt.org <mailto:Users at ovirt.org> 
 >>>>> http://lists.ovirt.org/mailman/listinfo/users <http://lists.ovirt.org/mailman/listinfo/users> 
 >>>>
 >>>>
 >>>
 >>> _______________________________________________
 >>> Users mailing list
 >>> Users at ovirt.org <mailto:Users at ovirt.org> 
 >>> http://lists.ovirt.org/mailman/listinfo/users <http://lists.ovirt.org/mailman/listinfo/users> 
 >>>
 >>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ovirt.org/pipermail/users/attachments/20160823/7d56e02e/attachment-0001.html>


More information about the Users mailing list