Opennebula 5 - Xen and Openvswitch issues

Hello

  • 1 and Congrats to everyone on the opennebula 5.0 release

i installed a fresh opennebula 5.0 wizard setup and have setup xen using the add-ons
i want to use the openvswitch because of the support for vlans amongst other functions (eventually using a lot of SDN features)
now when i insantiate a vm, i see a boot failure and the following error on the logs
[[Z0][VMM][I]: No tap device found for nic 0

it goes into boot_failure state but when i login to the node and startup the vm using xl creatte, it comes up normally and i am able to recover the state to running

so my obvious question is whether the ovswitch implementation in opennebula 5 is specific to kvm in this case or am i doing something obviously wrong here ?

we have a 4.12 setup with about a 100 hosts on opennebula+xen+ovswitch and they work very well. We would like to test/do an upgrade to nebula 5.0 from this…


Imran

Hi

We can try to debug it.

1.- The VM failed to boot so you should have it in failure and all the file
there
2.- Go to the host and look for the VM directory
/var/lib/one/datastores/<system_ds_id>/<vm_id>
3.- Take a look to the deployment.0 file. Does it look correctly? trying to
starting the domian through xen gives you more information? Do you have the
ovswicth up in the host prior to the deployment?

Cheers

Ruben

Hi Ruben

Yes. I am able to see the deployment files as deployment.0 and its normal

name = 'one-14' memory = '1024' vcpus = '1' bootloader = 'pygrub' disk = [ 'raw:/var/lib/one//datastores/0/14/disk.0,hda,w', 'raw:/var/lib/one//datastores/0/14/disk.1,hdb,r', ] vif = [ ' mac=02:00:2b:fc:58:96,ip=,bridge=public0', ] vfb = ['type=vnc,vnclisten=0.0.0.0,vncunused=0,vncdisplay=14,vncpasswd='] I am able able to start the vms manually using xl create /var/lib/one/datastores/0/13/deployment.0

Once the vm is up, i am able to access the box as well and then use the recover option to mark the vm as success

But the problem is if i subsequently poweroff and poweron from sunstone , the error come in ( [[Z0][VMM][I]: No tap device found for nic 0 )

Pls let me know if any further debug logs or info is needed…

Imran

Hi thanks,

Seems to be a bug with the openvswitch drivers looking for the Xen devices.
We’ll look into it.

THANKS!

One more, thing can you be sure that the hosts have the last version of the
drivers.

onehost sync -f …

Make sure you have a nic-xen.rb in the /var/tmp/one/ directory in the Xen
hypervisors. The code is the same as for 4.12 so may it is missing…

Hi Ruben

Strangely, i am getting an error in doing onehost sync -f

[oneadmin@SPK-X-0140 ~]$ onehost sync -f * Adding spkx-0141 to upgrade [========================================] 1/1 spkx-0141 Failed to update the following hosts: * spkx-0141

Although i am able to see the latest files (nic-xen.rb) synced over to the hyervisor


[root@SPK-X-0141 one]# pwd
/var/tmp/one
[root@SPK-X-0141 one]# find * | grep nic-xen
vnm/vnmmad-load.d/nic-xen.rb

Imran

Pls let me know if this is a bug or is this something of a config issue ? Pls do let me know in case more info is required to troubleshoot this further


Imran

am able to re-produce the same issue in the latest 5.2 setup as well with xen and openvswitch…

We patched the nic.rb to detect which hypervisor is it and then attach the device to openvswitch