On 03.10.11 15:20, Daniel P. Berrange wrote:
On Mon, Oct 03, 2011 at 04:09:42PM +0200, Shahar Havivi wrote:
> On 03.10.11 12:24, Daniel P. Berrange wrote:
> > On Mon, Oct 03, 2011 at 01:09:31PM +0200, Shahar Havivi wrote:
> > > Hi,
> > >
> > > I am trying to run a VM (via VDSM) with hugepages and I am getting this
error:
> > > "virNetClientProgramDispatchError:170 : internal error hugetlbfs
filesystem is not mounted"
> > >
> > >
> > > hugepages is mounted with the proper permissions (36:36) in case of vdsm
user,
> > > this code use to work..., now I am at version:
libvirt-0.9.4-12.el6.x86_64
> > >
> > > output of few commands:
> > >
> > > # tree /dev/hugepages/
> > > /dev/hugepages/
> > > `-- libvirt
> > > `-- qemu
> > >
> > > # ll -n /dev/hugepages/libvirt/
> > > total 0
> > > drwxr-xr-x 2 36 36 0 Oct 3 12:46 qemu/
> >
> > The user ID '36:36' looks a little bit strange. Did you set that
> this is the userid that vdsm use
> > yourself ? When libvirtd starts up, it should detect the mount
> > point and then create the $MOUNT/libvirt/qemu directory itself,
> > setting up the correct userid/groupid which i'd expect to be
> > 107:107 on Fedora/Rhel
> >
> > In addition, was the hugetlbfs mount point actually mounted at
> > the time libvirtd was started up ? If you mount hugetlbfs after
> > starting libvirtd, you will need to restart it to enable hugetlbs
> > in libvirtd.
> No, it was not mounted prior to libvirt startup
Looking at the libvirtd code, AFAICT, the only way libvirtd can generate
that error message is if the hugetlbfs mount point was not present when
it started. What does
# grep hugepages /proc/mounts
show on the host in question ?
Does restarting libvirtd make it work
Daniel
Yes thanks,
When mounting the hugepages before libvirt start it works fine.
Thank you,
Shahar Havivi.