Ask Your Question
0

SR-IOV and Linuxbridge, 2 compute nodes?

asked 2017-05-17 01:31:45 -0500

Diavel gravatar image

Hello guys.. has anyone bumped into the same problem here?

I am setting up an environment with 1 controller + 2 compute nodes here, using Ubuntu 16.04.2 LTS and Openstack Ocata. I installed everything from repos..

apt install software-properties-common
add-apt-repository cloud-archive:ocata

I followed this guide: https://docs.openstack.org/ocata/install-guide-ubuntu/ (https://docs.openstack.org/ocata/inst...) and https://docs.openstack.org/ocata/networking-guide/config-sriov.html#config-sriov (https://docs.openstack.org/ocata/netw...) even tried directly attaching PCI device to VM but can't seem to make this working.

So basically I have 1 network as a MGMT (iface "eno1"), which has access to the internet and second one (enp10s0f0) which is only connected between the three PCs via switch and I'm using neutron-linuxbridge-agent + neutron-sriov-agent.

I've created a flat provider linuxbridge network using the second enp10s0f0 iface and I am able to launch instances using this linuxbridge net. I just do: openstack server create --flavor m1.medium --image trusty --security-group default --nic net-id=$NET --key-name mykey $vmName --availability-zone nova:compute1 which works on both compute nodes.

What doesn't work is when I want to want to pass a sriov port onto the second compute node...

openstack port create --network $provider_NET --vnic-type direct --host compute2 sriov_port2

openstack server create --flavor m1.medium --image trusty --security-group default --nic port-id=$sriov_port2_ID --nic net-id=$provider_NET --key-name mykey $vmName --availability-zone nova:compute2

Weird thing is that I am able to launch instance on the first compute node doing exactly the same thing.
What I am always getting on the second compute node is: Insufficient compute resources: Claim pci failed..

I have created the VFs (virtual functions) on both compute nodes the same way. Both having same NIC.

Am I missing something here? :(
Thanks.

edit retag flag offensive close merge delete

1 answer

Sort by ยป oldest newest most voted
0

answered 2017-05-24 02:09:31 -0500

Diavel gravatar image

I actually found out solution myself.. What had to be changed was /etc/nova/nova.conf and added:

vif_plugging_timeout=20
vif_plugging_is_fatal=False

Obviously while attaching VF into the VM it timed out and because of that the VM was forced into the ERROR state and show Claim pci failed.

Anyway it works now, only the VF won't get an IP address, because I am using different iface than the one that's used for linuxbridge provider net.

edit flag offensive delete link more

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Get to know Ask OpenStack

Resources for moderators

Question Tools

1 follower

Stats

Asked: 2017-05-17 01:31:45 -0500

Seen: 409 times

Last updated: May 24 '17