[openstack-dev] [nova][neutron] numa aware vswitch

Stephen Finucane sfinucan at redhat.com
Mon Aug 27 09:36:54 UTC 2018


On Mon, 2018-08-27 at 10:24 +0100, Sean Mooney wrote:
> 
> 
> On Mon 27 Aug 2018, 04:20 Guo, Ruijing, <ruijing.guo at intel.com> wrote:
> > Hi, Stephen,
> > 
> > After setting flavor, VM was created in node 0 (expect in node1). How to debug it?
> > 
> > Nova.conf
> > [neutron]
> > physnets = physnet0,physnet1
> > 
> > [neutron_physnet_physnet1]
> > numa_nodes = 1
> 
> Have you enabled the numa topology filter its off by default and without it the numa aware vswitch code is disabled.

Yeah, make sure this is enabled. You should turn on debug-level logging
as this will give you additional information about how things are being
scheduled. Also, is this a new deployment? If not, you're going to need
to upgrade and restart all the nova-* services since there are object
changes which will need to be propagated.

Stephen

> > openstack network create net1 --external --provider-network-type=vlan --provider-physical-network=physnet1 --provider-segment=200
> > ...
> > openstack server create --flavor 1 --image=cirros-0.3.5-x86_64-disk --nic net-id=net1 vm1
> > 
> > <cputune>
> >     <shares>1024</shares>
> >     <vcpupin vcpu='0' cpuset='0-7,16-23'/>
> >     <emulatorpin cpuset='0-7,16-23'/>
> >   </cputune>
> > 
> > available: 2 nodes (0-1)
> > node 0 cpus: 0 1 2 3 4 5 6 7 16 17 18 19 20 21 22 23
> > node 0 size: 64412 MB
> > node 0 free: 47658 MB
> > node 1 cpus: 8 9 10 11 12 13 14 15 24 25 26 27 28 29 30 31
> > node 1 size: 64502 MB
> > node 1 free: 44945 MB
> > node distances:
> > node   0   1 
> >   0:  10  21 
> >   1:  21  10
> > 
> > Thanks,
> > -Ruijing
> > 
> > -----Original Message-----
> > From: Stephen Finucane [mailto:sfinucan at redhat.com] 
> > Sent: Saturday, August 25, 2018 12:15 AM
> > To: OpenStack Development Mailing List (not for usage questions) <openstack-dev at lists.openstack.org>
> > Subject: Re: [openstack-dev] [nova][neutron] numa aware vswitch
> > 
> > On Fri, 2018-08-24 at 09:13 -0500, Matt Riedemann wrote:
> > > On 8/24/2018 8:58 AM, Stephen Finucane wrote:
> > > > Using this won't add a NUMA topology - it'll just control how any 
> > > > topology present will be mapped to the guest. You need to enable 
> > > > dedicated CPUs or a explicitly request a NUMA topology for this to 
> > > > work.
> > > > 
> > > > openstack flavor set --property hw:numa_nodes=1 1
> > > > 
> > > > <or>
> > > > 
> > > > openstack flavor set --property hw:cpu_policy=dedicated 1
> > > > 
> > > > 
> > > > This is perhaps something that we could change in the future, though 
> > > > I haven't given it much thought yet.
> > > 
> > > Looks like the admin guide [1] should be updated to at least refer to 
> > > the flavor user guide on setting up these types of flavors?
> > > 
> > > [1] 
> > > https://docs.openstack.org/nova/latest/admin/networking.html#numa-affi
> > > nity
> > 
> > Good idea.
> > 
> > https://review.openstack.org/596393
> > 
> > Stephen




More information about the OpenStack-dev mailing list