11 Installing the OpenStack Nodes #

Log in as user root to a machine in the Ceph cluster and generate keyring files for cinder users. Optionally, you can generate keyring files for the glance users (only needed when using glance with Ceph/Rados). The keyring file that will be generated for cinder will also be used on the Compute Nodes. To do so, you need to specify pool names and user names for both services. The default names are:

Make a note of user and pool names in case you do not use the default values. You will need this information later, when deploying glance and cinder.

Warning: Automatic Changes to the Cluster

If you decide to use the admin keyring file to connect the external Ceph cluster, be aware that after Crowbar discovers this admin keyring, it will create client keyring files, pools, and capabilities needed to run glance, cinder, or nova integration.

If you have access to the admin keyring file and agree that automatic changes will be done to the cluster as described above, copy it together with the Ceph configuration file to the Administration Server. If you cannot access this file, create a keyring:

1. When you can access the admin keyring file ceph.client.admin.keyring, copy it together with ceph.conf (both files are usually located in /etc/ceph) to a temporary location on the Administration Server, for example /root/tmp/.

2. If you cannot access the admin keyring file create a new keyring file with the following commands. Re-run the commands for glance, too, if needed. First create a key:

   ceph auth get-or-create-key client.USERNAME mon "allow r" \
   osd 'allow class-read object_prefix rbd_children, allow rwx \
   pool=POOLNAME'

   Replace USERNAME and POOLNAME with the respective values.

   Now use the key to generate the keyring file /etc/ceph/ceph.client.USERNAME.keyring:

   ceph-authtool \
   /etc/ceph/ceph.client.USERNAME.keyring \
   --create-keyring --name=client.USERNAME> \
   --add-key=KEY

   Replace USERNAME with the respective value.

   Copy the Ceph configuration file ceph.conf (usually located in /etc/ceph) and the keyring file(s) generated above to a temporary location on the Administration Server, for example /root/tmp/.

Log in to the Crowbar Web interface and check whether the nodes which should have access to the Ceph cluster already have an IP address from the storage network. Do so by going to the Dashboard and clicking the node name. An IP address should be listed for storage. Make a note of the Full name of each node that has no storage network IP address.

Log in to the Administration Server as user root and run the following command for all nodes you noted down in the previous step:

crowbar network allocate_ip "default" NODE "storage" "host"
chef-client

NODE needs to be replaced by the node's name.

After executing the command in the previous step for all affected nodes, run the command chef-client on the Administration Server.

Log in to each affected node as user root. See How can I log in to a node as root? for instructions. On each node, do the following:

1. Manually install nova, cinder (if using cinder) and/or glance (if using glance) packages with the following commands:

   zypper in openstack-glance
   zypper in openstack-cinder
   zypper in openstack-nova

2. Copy the ceph.conf file from the Administration Server to /etc/ceph:

   mkdir -p /etc/ceph
   scp root@admin:/root/tmp/ceph.conf /etc/ceph
   chmod 664 /etc/ceph/ceph.conf

3. Copy the keyring file(s) to /etc/ceph. The exact process depends on whether you have copied the admin keyring file or whether you have created your own keyrings:

   1. If you have copied the admin keyring file, run the following command on the Control Node(s) on which cinder and glance will be deployed, and on all KVM Compute Nodes:

      scp root@admin:/root/tmp/ceph.client.admin.keyring /etc/ceph
      chmod 640 /etc/ceph/ceph.client.admin.keyring

   2. If you have created you own keyrings, run the following command on the Control Node on which cinder will be deployed, and on all KVM Compute Nodes to copy the cinder keyring:

      scp root@admin:/root/tmp/ceph.client.cinder.keyring /etc/ceph
      chmod 640 /etc/ceph/ceph.client.cinder.keyring

      On Control Node on which cinder will be deployed run the following command to update file ownership:

      chown root.cinder /etc/ceph/ceph.client.cinder.keyring

      On KVM Compute Nodes run the following command to update file ownership:

      chown root.nova /etc/ceph/ceph.client.cinder.keyring

      Now copy the glance keyring to the Control Node on which glance will be deployed:

      scp root@admin:/root/tmp/ceph.client.glance.keyring /etc/ceph
      chmod 640 /etc/ceph/ceph.client.glance.keyring
      chown root.glance /etc/ceph/ceph.client.glance.keyring