Adelina's profile - activity

Could you please paste me the logs from both nova and cinder?

Can you please post nova-compute logs and cinder logs ? paste.openstack.org should be fine. Thank you, Adelina

Hello,

This looks to me like you are using different versions of OpenStack on your controller vs your compute node.

The compute node seems to be on master, what are you using for the controller?

Hello,

Once you have created your switch, which in your case does not have a AllowManagementOS set to True, you cand do something like this:

Set-VMSwitch -AllowManagementOS $True switch-name

This is to enable AllowManagementOS on the switch. Then, to enable vlan you need to use the following command:

   Get-VMNetworkAdapter -SwitchName switch-name -ManagementOS | Set-VMNetworkAdapterVlan -Access -VlanId VlanID

Hello Hugo,

Can you please give some more details about your environment? What version of OVS version are you using? Did you enable the "OVS Hyper-V virtual switch forwarding extension" on the switch you are using? Can also describe step by step what you are doing?

We didn't manage to reproduce your problem. We spawned vms with multiple NICs and multiple vms at a time, all was fine.

First of all you have to make sure you use the compute hyperv driver. You can check this by looking in c:\ProgramFiles (x86)\CloudbaseSolutions\Openstack\Nova\etc\nova.conf and see if compute_driver = hyperv.nova.driver.HyperVDriver

In order to use RemoteFX you have to create a new flavor in Openstack with the following extra-spec: {"hyperv:remotefx": "1920x1200,1"}. In this case, it means that your vm will have remotefx enabled with 1920x1200 resolution and 1 monitor. Eg:

nova flavor-create m1.remotefx 111 4096 20 2 nova flavor-key 11 set hyperv:remotefx="1920x1200,1"

By doing this, you create a new flavor for remotefx enabled machines.

After this, you just have to boot a new machine with this flavor and it will have a remotefx video adapter.

Note that the valid resolution values for 2012R2 are: 1024x768, 1280 ×1024, 1600 ×1200, 1920 ×1200 and the maximum number of monitors is 4 for resolutions up to 1280x1024, 3 for 1600x1200 and 2 for 1920x1200.

Thank you,

Adelina

Support for OVS neutron agent was added in Kilo, there is no support for Juno. There are a number of patches both on nova and on neutron that enable the neutron agent to work on Hyper-V compute nodes, backporting those to your Juno env is not a trivial task.

This is odd. Could you please tell me what branch are you using? Also, if you can paste the full logs from c:\Openstack\Logs\nova-compute.log from both hosts, that would be helpful. Please use pastebin or paste.openstack.org

EDIT:

Sorry for answering so late, my original response didn’t get posted from my phone for some reason.

I tried to reproduce, but couldn’t find any issue. I tried with both Kilo and Liberty branches, both work fine. You didn’t said which branch you were using, but judging by the line numbers in the trace I assumed Kilo.

You posted the logs from both Hyper-V Nodes, but unfortunately there is not much information there. I suspect that this exception is masking the real problem, even so, it is odd that there is no Hyper-V error in the logs.

Given that I can confirm that live-migration works, I suspect that the problem is related to the configuration of the Hyper-V nodes. One thing you can try is to start a migration and immediately after it fails, check to see if there is any error with Event Viewer on both Hyper-V Hosts. You should look under “Application and Services\Microsoft\Windows\Hyper-V VMMS”. Also, please set your log level to debug=true in the nova.conf file on both hosts, that should give us a clearer picture of what is happening.

There is no public release of vNext tp4 I can use so unfortunately I cannot help you by reproducing your specific environment. Again, my intuition is that the nested virtualization might be the problem. We tested Nano as a compute node on tp3.

This is the link for the gerrit review: https://review.openstack.org/#/c/242952/ Once it merges on master we will backport to liberty and kilo. Either way, the fix is extremely trivial so you can modify the source file yourself (until it merges) if your environment is for testing purposes.

It is a bug, we are working on a fix right now. The problem is that that attribute is named localnetworkvswitch but here it is called as localnetwork_vswitch.

I'll try to replicate your environment to see if I get issues. My intuition is that the nested virtualization might cause issues.

I edited my original answer. If you can give me those details we can start figuring out exactly what went wrong in your environment and how to fix it.

Sure you can use Nano as a compute Node. Starting with Liberty we release a zip archive containing the python environment and all you need to run nova-compute on Nano server. You can download the zip at this link

All you need to do is make a configuration file for nova-compute and neutron-hyperv-agent. This is a sample configuration for nova-compute and this one is for neutron-hyperv-agent. Modify passwords and ips to match your specific environment. You can start the services by using:

PS C:\HyperVNovaCompute_Liberty_12_0_test\Python> .\python.exe .\Scripts\nova-compute-script.py --config-file ..\etc\nova.conf

And for neutron-hyperv-agent:

C:\HyperVNovaCompute_Liberty_12_0_test\Python> .\python.exe .\Scripts\neutron-hyperv-agent-script.py --config-file ..\etc\neutron_hyperv_agent.conf

Our compute driver works very well with devstack. What exactly has been your problem?

EDIT:

Before we can start debugging your environment can you please provide me some details?

1. Do you use physical machines for the deployment or virtual?
2. What config file did you use to deploy devstack? Does the devstack deployment work or are there errors. If so, can you please provide logs?
3. How did you install nano? Does it have the Hyperv role enabled? Also, can you provide the logs to see what happened when you tried to start nova? The logs should be under c:\OpenStack\Log.

Could you please post the output of the following commands:

neutron net-show your-net-id

and on Hyper-V Server

get-VMSwitch

get-VMNetworkAdapterVlan -VMName name-of-vm

Also, you say that they are not getting dhcp addresses, what happens if you set the address manually? Does it work then?

At the end of this page there is a guide for troubleshooting that has a section refering to Vlan filtering by the NIC driver. Please check if that is the case.

Could you please post the logs for the HyperV agent?

You can find the logs in c:\Openstack\Logs\neutron-hyperv-agent.log

I would suggest using pastebin or paste.openstack.org rather.

Thank you! Adelina

The error occurs before nova begins to start logging. You can start nova-compute in a powershell session with

C:\Program Files (x86)\Cloudbase Solutions\OpenStack\Nova\Python27\Scripts\nova-compute.exe --config-file c:\Program Files (x86)\Cloudbase Solutions\OpenStack\Nova\etc\nova.conf

I think it is a credential error but to be sure, running the above command will output the error. Could you please copy paste the result here?

Could you please post the logs? The log dir is c:\Openstack\Log

To run the service "manually", open cmd and run: In case you use Kilo branch:

python "C:\Program Files (x86)\Cloudbase Solutions\OpenStack\Python27\Scripts\cinder-volume-script.py" --config-file "C:\ProgramFiles(x86)\CloudbaseSolutions\OpenStack\etc\cinder\cinder.conf"

In case you use stable/Juno branch:

python "C:\Program Files (x86)\Cloudbase Solutions\OpenStack\Python27\Scripts\cinder-volume" --config-file "C:\Program Files (x86)\Cloudbase Solutions\OpenStack\etc\cinder\cinder.conf"

I can't really make out much from the log from event viewer. Isn't there any log file in c:\Openstack\Log or the path you configured when you ran the installer ? If not, you could try to run the service from cmd and check the output there.

Hello Partha,

It would be great of you could post the log so we can figure out what the problem is. The log should be located in c:\Openstack\Log . Either copy and paste it here or if it's to large, use http://paste.openstack.org/.

Thanks

I've been looking trough the nova output and sadly I don't see much that can help there.Can you give me the nova-compute logs? Also, is your image format qcow2 ?

Could you give the nova log? Paste it in paste.openstack.org or something similar. Also, what kind of flavor are you using for the instance? If you can provide some details about your compute node(s) that would also be of use. Thanks!