Description of problem: when using the gluster backend with fuse, snapshotting an "in-use" volume fails with the following in volume.log 2014-02-23 00:52:40.644 882 ERROR cinder.volume.drivers.glusterfs [req-e794363d-a226-41ee-ba88-8a1c4969995f 538f6959240b4d4db75cb3569868c0bf 482602d8975f47d79ec6a8cb70fe8f62] Call to Nova to create snapshot failed 2014-02-23 00:52:40.644 882 ERROR cinder.volume.drivers.glusterfs [req-e794363d-a226-41ee-ba88-8a1c4969995f 538f6959240b4d4db75cb3569868c0bf 482602d8975f47d79ec6a8cb70fe8f62] Policy doesn't allow compute_extension:os-assisted-volume-snapshots:create to be performed. (HTTP 403) (Request-ID: req-26dd4f46-c961-4227-ac07-de685f9938ed) 2014-02-23 00:52:40.644 882 TRACE cinder.volume.drivers.glusterfs Traceback (most recent call last): 2014-02-23 00:52:40.644 882 TRACE cinder.volume.drivers.glusterfs File "/usr/lib/python2.6/site-packages/cinder/volume/drivers/glusterfs.py", line 408, in _create_snapshot 2014-02-23 00:52:40.644 882 TRACE cinder.volume.drivers.glusterfs connection_info) 2014-02-23 00:52:40.644 882 TRACE cinder.volume.drivers.glusterfs File "/usr/lib/python2.6/site-packages/cinder/compute/nova.py", line 130, in create_volume_snapshot 2014-02-23 00:52:40.644 882 TRACE cinder.volume.drivers.glusterfs create_info=create_info) 2014-02-23 00:52:40.644 882 TRACE cinder.volume.drivers.glusterfs File "/usr/lib/python2.6/site-packages/novaclient/v1_1/contrib/assisted_volume_snapshots.py", line 41, in create 2014-02-23 00:52:40.644 882 TRACE cinder.volume.drivers.glusterfs return self._create('/os-assisted-volume-snapshots', body, 'snapshot') 2014-02-23 00:52:40.644 882 TRACE cinder.volume.drivers.glusterfs File "/usr/lib/python2.6/site-packages/novaclient/base.py", line 145, in _create 2014-02-23 00:52:40.644 882 TRACE cinder.volume.drivers.glusterfs _resp, body = self.api.client.post(url, body=body) 2014-02-23 00:52:40.644 882 TRACE cinder.volume.drivers.glusterfs File "/usr/lib/python2.6/site-packages/novaclient/client.py", line 232, in post 2014-02-23 00:52:40.644 882 TRACE cinder.volume.drivers.glusterfs return self._cs_request(url, 'POST', **kwargs) 2014-02-23 00:52:40.644 882 TRACE cinder.volume.drivers.glusterfs File "/usr/lib/python2.6/site-packages/novaclient/client.py", line 213, in _cs_request 2014-02-23 00:52:40.644 882 TRACE cinder.volume.drivers.glusterfs **kwargs) 2014-02-23 00:52:40.644 882 TRACE cinder.volume.drivers.glusterfs File "/usr/lib/python2.6/site-packages/novaclient/client.py", line 195, in _time_request 2014-02-23 00:52:40.644 882 TRACE cinder.volume.drivers.glusterfs resp, body = self.request(url, method, **kwargs) 2014-02-23 00:52:40.644 882 TRACE cinder.volume.drivers.glusterfs File "/usr/lib/python2.6/site-packages/novaclient/client.py", line 189, in request 2014-02-23 00:52:40.644 882 TRACE cinder.volume.drivers.glusterfs raise exceptions.from_response(resp, body, url, method) 2014-02-23 00:52:40.644 882 TRACE cinder.volume.drivers.glusterfs Forbidden: Policy doesn't allow compute_extension:os-assisted-volume-snapshots:create to be performed. (HTTP 403) (Request-ID: req-26dd4f46-c961-4227-ac07-de685f9938ed) Version-Release number of selected component (if applicable): openstack-cinder-2013.2.2-1.el6ost.noarch Steps to Reproduce: 1. configure cinder with gluster 2. create a volume and boot an instance from it 3. try to create a volume snapshot
This is a deployment issue. Nova must have an admin API endpoint configured. (Or for testing it is also possible to change Nova's policy.json to not require admin_api for this extension.)
Giulio: How was this deployed? The main thing here is to make sure that our documentation and deployment mechanisms deploy a Nova admin API endpoint.
this was installed using packstack
As it sounds like a Packstack/Puppet bug in deploying Cinder with Nova Admin API Endpoint, moving to Packstack.
Eric, is there somewhere an wiki about admin API deployment? The only thing I was able to find using my google-fu was load of blueprints. The second question assuming that Nova admin API is a new service is: is this service available in Havana also or is it new for Icehouce?
hi Martin, given this will have to be fixed in foreman, shall we move this to openstack-puppet-modules ?
(In reply to Martin Magr from comment #7) Deploying the admin API is actually not a requirement I think -- I'll look into this again from the Cinder side.
This is not a Z-stream material and/or test blocker.