On 11/11/2012 11:59 PM, Dennis Jacobfeuerborn wrote:
On 11/09/2012 11:33 AM, Pádraig Brady wrote:
On 11/09/2012 02:07 AM, Dennis Jacobfeuerborn wrote:
On 11/09/2012 01:47 AM, Pádraig Brady wrote:
On 11/09/2012 12:11 AM, Dennis Jacobfeuerborn wrote:
On 11/09/2012 12:55 AM, Pádraig Brady wrote: [SNIP]
That was the bug fixed (supposedly) by the openstack-utils update I referenced above. What has probably happened is that you were just ahead of your fedora mirror (which was one of the reasons we were using a side repo for the test days). Anyway you can update openstack-demo-install manually like:
rpm -Uvh http://kojipkgs.fedoraproject.org//packages/openstack-utils/2012.2/6.fc18/no...
I'm going to try this right away, thanks.
Note the new openstack-utils now does no keystone<->quantum setup at all, so that horizon is happy by default. So I've just now updated: https://fedoraproject.org/wiki/QA:Testcase_Quantum_V2 to add the required quantum keystone config steps.
openstack-demo-install can't really be called again and spits out lots of errors when I do so (it tries to recreate users and doesn't rewrite e.g. /etc/nova/nova.conf) so it seems wiping the data manually is the only way to go.
Anyway after doing so with the new openstack-utils rpm I now get a nova-network based setup and after creating a network from the shell and starting an instance I can successfully ssh into it using the internal IP from the host.
Good stuff.
I was unable to attach a floating ip though. I can add IPs to the pool ok but when I try to allocate one of the IPs for the VM with "nova floating-ip-create" I get a failure. The following appears in the nova api log: ... 2012-11-09 02:50:03 TRACE nova.api.openstack File "/usr/lib/python2.7/site-packages/nova/api/openstack/compute/contrib/floating_ips.py",
line 171, in create 2012-11-09 02:50:03 TRACE nova.api.openstack address = self.network_api.allocate_floating_ip(context, pool) 2012-11-09 02:50:03 TRACE nova.api.openstack File "/usr/lib/python2.7/site-packages/nova/network/quantumv2/api.py", line 322, in allocate_floating_ip 2012-11-09 02:50:03 TRACE nova.api.openstack raise NotImplementedError() 2012-11-09 02:50:03 TRACE nova.api.openstack NotImplementedError
So the mention of quantumv2 above is surprising. I'm unsure if you're using quantum or nova networking, but it seems like this run you're using nova networking but may have left stale config pertaining to quantum in /etc/nova/nova.conf. Please check the 'network_api_class' option, and remove it if you are in fact using nova networking.
If you are using quantum here, there there are quantum specific commands for setting up floating IPs at the end of: https://fedoraproject.org/wiki/QA:Testcase_Quantum_V2#Setup
When I tried again the next day I could assign a floating IP without problems. Maybe this was a caching issue or my shell env wasn't setup properly.
With that I can now start a single node setup and launch virtual machines with proper external connectivity.
I noticed that the quota information on the overview page in horizon is not displayed properly and could trace this back to the following patch: https://review.openstack.org/#/c/14379/ With this patch the quota is displayed correctly.
Cool. We'll look at getting that merged to stable/folsom upstream, and then into our packages.
Next I tried the following instructions to add a second node: https://fedoraproject.org/wiki/QA:Testcase_separate_OpenStack_compute_node
The page referenced "openstack-config-set" which doesn't exist (anymore?). Apparently this is now replaced by the syntax "openstack-config --set". I changed the page accordingly.
Excellent, thanks.
After that I can start the compute service and see it in the service list on the controller:
[dennis@controller ~]$ sudo nova-manage service list Binary Host Zone Status State Updated_At nova-compute controller nova enabled :-) 2012-11-11 23:55:50 nova-cert controller nova enabled :-) 2012-11-11 23:55:55 nova-scheduler controller nova enabled :-) 2012-11-11 23:55:55 nova-network controller nova enabled :-) 2012-11-11 23:55:54 nova-consoleauth controller nova enabled :-) 2012-11-11 23:55:57 nova-console controller nova enabled :-) 2012-11-11 23:55:50 nova-compute node nova enabled :-) 2012-11-11 23:55:58
Notice the two compute entries on "controller" and "node".
The Problem is that I cannot launch instances on the second node. The compute log shows the following: TRACE nova.openstack.common.rpc.amqp ImageNotAuthorized: Not authorized for image f4a80926-fcf7-4758-8a19-ba8c37f3cd16.
Apparently the second node cannot retrieve the image from glance?
Hmm, seems like we need to explicitly config to use keystone on the compute node: Could you try this line I've added to the wiki:
sudo openstack-config --set /etc/nova/nova.conf DEFAULT auth_stragegy keystone
thanks, Pádraig.