Open Stack

Hi Andy,

Yes it probably does not matter if the variables are visible in other 
containers (you get a working system up and running anyway) but it is 
not the expected behavior when using limit_container_types. I did manage 
to get the variables limited to cinder_volume containers when I typed 
the configuration like this:

   XxX:
     ip: 172.22.5.9
     container_vars:
       cinder_storage_availability_zone: cinderAZ_1
       cinder_default_availability_zone: cinderAZ_1
       cinder_backends:
         limit_container_types: cinder_volume
         cinder_nfs:
           volume_backend_name: cinder_nfs
           volume_driver: cinder.volume.drivers.nfs.NfsDriver
           nfs_mount_options: 
"_netdev,auto,rw,intr,noatime,async,vers=3,proto=tcp,wsize=1048576,rsize=1048576,timeo=1200,actimeo=120"
       cinder_nfs_client:
         limit_container_types: cinder_volume
         nfs_shares_config: /etc/cinder/nfs_shares
         shares:
           - ip: "172.22.20.254"
             share: "/nfs/cinder/production"


However, I did not find any way to limit the glance nfs configuration to 
only the glance containers. It looks like this in 
openstack_user_config.yml.prod.example:

image_hosts:
   infra1:
     ip: 172.29.236.11
     container_vars:
       limit_container_types: glance
       glance_nfs_client:
         - server: "172.29.244.15"
           remote_path: "/images"
           local_path: "/var/lib/glance/images"
           type: "nfs"
           options: "_netdev,auto"

My workaround to the problem was to add the glance configuration to 
user_variables.yml instead, that way you don't get the inventory full 
with glance configuration on all containers. This is what I added to 
user_variables.yml:

glance_nfs_client:
   - server: "172.22.20.254"
     remote_path: "/nfs/glance/production"
     local_path: "/var/lib/glance/images"
     type: "nfs"
     options: 
"_netdev,auto,rw,intr,noatime,async,vers=3,proto=tcp,wsize=1048576,rsize=1048576,timeo=1200,actimeo=120"


So it sounds like you are aware of the problem and are working on a fix, 
that is great.
Thank you all for a very good work with openstack-ansible.

Regards,
Andreas



On 02/03/2017 05:27 PM, Andy McCrae wrote:
> Hi Andreas,
>
> The way you're doing it at the end looks correct - the docs are not 
> quite right on that one.
> The nfs_shares file will only get templated on the cinder_volumes 
> hosts, as will the nfs_shares_config option  - so in essence it 
> shouldn't matter that the var isn't limited to volume hosts.
>
> That said, there is a bug here that we've been working to fix - this 
> will work if you have one backend.
> Mainly it means there is no support for multiple backends, if one of 
> them is an NFS backend, and there is no support for multiple NFS backends.
> A patch has gone in to allow multiple NFS backends (and change the way 
> they're configured) - but it's not quite ready for release - so we're 
> working on a subsequent fix that will then be a part of newton 14.0.8 
> release for OSA.
>
> I'd be happy to keep you updated - if you'd like to drop into 
> #openstack-ansible on Freednode IRC there are a lot of active 
> operators/developers/deployers, we'd be happy to help further.
>
> But barring that I'll try add a reminder to drop an update here once 
> that's working.
>
> Andy
>
>
> On 3 February 2017 at 08:56, Andreas Vallin <andreas.vallin at it.uu.se 
> <mailto:andreas.vallin at it.uu.se>> wrote:
>
>     Hi!
>
>     I need to ask how do you correctly configure nfs to be used with
>     openstack-ansible newton (14.0.6). I think it is great that there
>     is an production example file that uses nfs for glance and cinder
>     (openstack_user_config.yml.pro
>     <http://openstack_user_config.yml.pro>d.example) but the cinder
>     config is not working for me.
>
>     I use this config for storage_hosts, only changing ip, and share
>     from production example:
>
>     storage_hosts:
>       XxXx:
>         ip: 172.22.5.9
>         container_vars:
>           cinder_backends:
>             limit_container_types: cinder_volume
>             cinder_nfs_client:
>               nfs_shares_config: /etc/cinder/nfs_shares
>               shares:
>                 - ip: "172.22.20.254"
>                   share: "/nfs/cinder/production"
>
>     And this is the failure when running os-cinder-install.yml
>
>     TASK [os_cinder : Add in cinder devices types]
>     *********************************
>     fatal: [XxXx_cinder_volumes_container-080139bd]: FAILED! =>
>     {"failed": true, "msg": "the field 'args' has an invalid value,
>     which appears to include a variable that is undefined. The error
>     was: 'dict object' has no attribute 'volume_backend_name'\n\nThe
>     error appears to have been in
>     '/etc/ansible/roles/os_cinder/tasks/cinder_backends.yml': line 30,
>     column 3, but may\nbe elsewhere in the file depending on the exact
>     syntax problem.\n\nThe offending line appears to be:\n\n\n- name:
>     Add in cinder devices types\n  ^ here\n"}
>
>     OK, so volume_backend_name is missing. The playbook runs if I add
>     volume_backend_name to the config like this:
>
>
>     storage_hosts:
>       XxXx:
>         ip: 172.22.5.9
>         container_vars:
>           cinder_backends:
>             limit_container_types: cinder_volume
>             cinder_nfs_client:
>               volume_backend_name: cinder_nfs
>               nfs_shares_config: /etc/cinder/nfs_shares
>               shares:
>                 - ip: "172.22.20.254"
>                   share: "/nfs/cinder/production"
>
>
>     But now there is no /etc/cinder/nfs_shares file in the
>     cinder-volumes-container so the nfs share will not be mounted.
>     This is because the "Create nfs shares export file" task in
>     cinder_post_install.yml doesn't see that cinder_nfs_client is
>     defined. You also get this in cinder.conf:
>
>     enabled_backends=cinder_nfs_client
>     # All given backend(s)
>     [cinder_nfs_client]
>     volume_backend_name=cinder_nfs
>     nfs_shares_config=/etc/cinder/nfs_shares
>     shares=[{u'ip': u'172.22.20.254', u'share':
>     u'/nfs/cinder/production'}]
>
>
>     This configuration works for me:
>
>     storage_hosts:
>       XxXx:
>         ip: 172.22.5.9
>         container_vars:
>           cinder_storage_availability_zone: cinderAZ_1
>           cinder_default_availability_zone: cinderAZ_1
>           limit_container_types: cinder_volume
>           cinder_backends:
>             cinder_nfs:
>               volume_backend_name: cinder_nfs
>               volume_driver: cinder.volume.drivers.nfs.NfsDriver
>               nfs_mount_options:
>     "_netdev,auto,rw,intr,noatime,async,vers=3,proto=tcp,wsize=1048576,rsize=1048576,timeo=1200,actimeo=120"
>           cinder_nfs_client:
>             nfs_shares_config: /etc/cinder/nfs_shares
>             shares:
>               - ip: "172.22.20.254"
>                 share: "/nfs/cinder/production"
>
>
>     BUT when I look in the inventory file (openstack_inventory.json)
>     it doesn't look like this configuration is limited to
>     cinder_volume containers even if "limit_container_types:
>     cinder_volume" is used. So now I feel it is time to ask how a
>     correct configuration should look like.
>
>     Regards,
>     Andreas
>
>
>     _______________________________________________
>     OpenStack-operators mailing list
>     OpenStack-operators at lists.openstack.org
>     <mailto:OpenStack-operators at lists.openstack.org>
>     http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators
>     <http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators>
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-operators/attachments/20170206/7b3c06ba/attachment.html>