Jump to contentJump to page navigation: previous page [access key p]/next page [access key n]
Applies to SUSE OpenStack Cloud 8

12 Alternative Configurations Edit source

In SUSE OpenStack Cloud 8 there are alternative configurations that we recommend for specific purposes.

12.1 Using a Dedicated Cloud Lifecycle Manager Node Edit source

All of the example configurations included host the Cloud Lifecycle Manager on the first Control Node. It is also possible to deploy this service on a dedicated node. One use case for wanting to run the dedicated Cloud Lifecycle Manager is to be able to test the deployment of different configurations without having to re-install the first server. Some administrators prefer the additional security of keeping all of the configuration data on a separate server from those that users of the cloud connect to (although all of the data can be encrypted and SSH keys can be password protected).

Here is a graphical representation of this setup:

12.1.1 Specifying a dedicated Cloud Lifecycle Manager in your input model Edit source

To specify a dedicated Cloud Lifecycle Manager in your input model, make the following edits to your configuration files.

Important
Important

The indentation of each of the input files is important and will cause errors if not done correctly. Use the existing content in each of these files as a reference when adding additional content for your Cloud Lifecycle Manager.

  • Update control_plane.yml to add the Cloud Lifecycle Manager.

  • Update server_roles.yml to add the Cloud Lifecycle Manager role.

  • Update net_interfaces.yml to add the interface definition for the Cloud Lifecycle Manager.

  • Create a disks_lifecycle_manager.yml file to define the disk layout for the Cloud Lifecycle Manager.

  • Update servers.yml to add the dedicated Cloud Lifecycle Manager node.

Control_plane.yml: The snippet below shows the addition of a single node cluster into the control plane to host the Cloud Lifecycle Manager service. Note that, in addition to adding the new cluster, you also have to remove the Cloud Lifecycle Manager component from the cluster1 in the examples:

  clusters:
     - name: cluster0
       cluster-prefix: c0
       server-role: LIFECYCLE-MANAGER-ROLE
       member-count: 1
       allocation-policy: strict
       service-components:
         - lifecycle-manager
         - ntp-client
     - name: cluster1
       cluster-prefix: c1
       server-role: CONTROLLER-ROLE
       member-count: 3
       allocation-policy: strict
       service-components:
         - lifecycle-manager
         - ntp-server
         - tempest

This specifies a single node of role LIFECYCLE-MANAGER-ROLE hosting the Cloud Lifecycle Manager.

Server_roles.yml: The snippet below shows the insertion of the new server roles definition:

   server-roles:

      - name: LIFECYCLE-MANAGER-ROLE
        interface-model: LIFECYCLE-MANAGER-INTERFACES
        disk-model: LIFECYCLE-MANAGER-DISKS

      - name: CONTROLLER-ROLE

This defines a new server role which references a new interface-model and disk-model to be used when configuring the server.

net-interfaces.yml: The snippet below shows the insertion of the network-interface info:

    - name: LIFECYCLE-MANAGER-INTERFACES
      network-interfaces:
        - name: BOND0
          device:
             name: bond0
          bond-data:
             options:
                 mode: active-backup
                 miimon: 200
                 primary: hed3
             provider: linux
             devices:
                 - name: hed3
                 - name: hed4
          network-groups:
             - MANAGEMENT

This assumes that the server uses the same physical networking layout as the other servers in the example.

disks_lifecycle_manager.yml: In the examples, disk-models are provided as separate files (this is just a convention, not a limitation) so the following should be added as a new file named disks_lifecycle_manager.yml:

---
   product:
      version: 2

   disk-models:
   - name: LIFECYCLE-MANAGER-DISKS
     # Disk model to be used for Cloud Lifecycle Managers nodes
     # /dev/sda_root is used as a volume group for /, /var/log and /var/crash
     # sda_root is a templated value to align with whatever partition is really used
     # This value is checked in os config and replaced by the partition actually used
     # on sda e.g. sda1 or sda5

     volume-groups:
       - name: ardana-vg
         physical-volumes:
           - /dev/sda_root

       logical-volumes:
       # The policy is not to consume 100% of the space of each volume group.
       # 5% should be left free for snapshots and to allow for some flexibility.
          - name: root
            size: 80%
            fstype: ext4
            mount: /
          - name: crash
            size: 15%
            mount: /var/crash
            fstype: ext4
            mkfs-opts: -O large_file
        consumer:
              name: os

Servers.yml: The snippet below shows the insertion of an additional server used for hosting the Cloud Lifecycle Manager. Provide the address information here for the server you are running on, that is, the node where you have installed the SUSE OpenStack Cloud ISO.

  servers:
     # NOTE: Addresses of servers need to be changed to match your environment.
     #
     #       Add additional servers as required

     #Lifecycle-manager
     - id: lifecycle-manager
       ip-addr: YOUR IP ADDRESS HERE
       role: LIFECYCLE-MANAGER-ROLE
       server-group: RACK1
       nic-mapping: HP-SL230-4PORT
       mac-addr: 8c:dc:d4:b5:c9:e0
       # ipmi information is not needed 

     # Controllers
     - id: controller1
       ip-addr: 192.168.10.3
       role: CONTROLLER-ROLE
Important
Important

With a stand-alone deployer, the OpenStack CLI and other clients will not be installed automatically. You need to install OpenStack clients to get the desired OpenStack capabilities. For more information and installation instructions, consult Book “Installing with Cloud Lifecycle Manager”, Chapter 28 “Installing OpenStack Clients”.

12.2 Configuring SUSE OpenStack Cloud without DVR Edit source

By default in the KVM model, the Neutron service utilizes distributed routing (DVR). This is the recommended setup because it allows for high availability. However, if you would like to disable this feature, here are the steps to achieve this.

On your Cloud Lifecycle Manager, make the following changes:

  1. In the ~/openstack/my_cloud/config/neutron/neutron.conf.j2 file, change the line below from:

    router_distributed = {{ router_distributed }}

    to:

    router_distributed = False
  2. In the ~/openstack/my_cloud/config/neutron/ml2_conf.ini.j2 file, change the line below from:

    enable_distributed_routing = True

    to:

    enable_distributed_routing = False
  3. In the ~/openstack/my_cloud/config/neutron/l3_agent.ini.j2 file, change the line below from:

    agent_mode = {{ neutron_l3_agent_mode }}

    to:

    agent_mode = legacy
  4. In the ~/openstack/my_cloud/definition/data/control_plane.yml file, remove the following values from the Compute resource service-components list:

    - neutron-l3-agent
       - neutron-metadata-agent
    Warning
    Warning

    If you fail to remove the above values from the Compute resource service-components list from file ~/openstack/my_cloud/definition/data/control_plane.yml, you will end up with routers (non_DVR routers) being deployed in the compute host, even though the lifecycle manager is configured for non_distributed routers.

  5. Commit your changes to your local git repository:

    ardana > cd ~/openstack/ardana/ansible
    ardana > git add -A
    ardana > git commit -m "My config or other commit message"
  6. Run the configuration processor:

    ardana > cd ~/openstack/ardana/ansible
    ardana > ansible-playbook -i hosts/localhost config-processor-run.yml
  7. Run the ready deployment playbook:

    ardana > cd ~/openstack/ardana/ansible
    ardana > ansible-playbook -i hosts/localhost ready-deployment.yml
  8. Continue installation. More information on cloud deployments are available in the Book “Installing with Cloud Lifecycle Manager”, Chapter 7 “Overview”

12.3 Configuring SUSE OpenStack Cloud with Provider VLANs and Physical Routers Only Edit source

Another option for configuring Neutron is to use provider VLANs and physical routers only, here are the steps to achieve this.

On your Cloud Lifecycle Manager, make the following changes:

  1. In the ~/openstack/my_cloud/config/neutron/neutron.conf.j2 file, change the line below from:

    router_distributed = {{ router_distributed }}

    to:

    router_distributed = False
  2. In the ~/openstack/my_cloud/config/neutron/ml2_conf.ini.j2 file, change the line below from:

    enable_distributed_routing = True

    to:

    enable_distributed_routing = False
  3. In the ~/openstack/my_cloud/config/neutron/dhcp_agent.ini.j2 file, change the line below from:

    enable_isolated_metadata = {{ neutron_enable_isolated_metadata }}

    to:

    enable_isolated_metadata = True
  4. In the ~/openstack/my_cloud/definition/data/control_plane.yml file, remove the following values from the Compute resource service-components list:

    - neutron-l3-agent
      - neutron-metadata-agent

12.4 Considerations When Installing Two Systems on One Subnet Edit source

If you wish to install two separate SUSE OpenStack Cloud 8 systems using a single subnet, you will need to consider the following notes.

The ip_cluster service includes the keepalived daemon which maintains virtual IPs (VIPs) on cluster nodes. In order to maintain VIPs, it communicates between cluster nodes over the VRRP protocol.

A VRRP virtual routerid identifies a particular VRRP cluster and must be unique for a subnet. If you have two VRRP clusters with the same virtual routerid, causing a clash of VRRP traffic, the VIPs are unlikely to be up or pingable and you are likely to get the following signature in your /etc/keepalived/keepalived.log:

Dec 16 15:43:43 ardana-cp1-c1-m1-mgmt Keepalived_vrrp[2218]: ip address
  associated with VRID not present in received packet : 10.2.1.11
Dec 16 15:43:43 ardana-cp1-c1-m1-mgmt Keepalived_vrrp[2218]: one or more VIP
  associated with VRID mismatch actual MASTER advert
Dec 16 15:43:43 ardana-cp1-c1-m1-mgmt Keepalived_vrrp[2218]: bogus VRRP packet
  received on br-bond0 !!!
Dec 16 15:43:43 ardana-cp1-c1-m1-mgmt Keepalived_vrrp[2218]: VRRP_Instance(VI_2)
  ignoring received advertisment...

To resolve this issue, our recommendation is to install your separate SUSE OpenStack Cloud 8 systems with VRRP traffic on different subnets.

If this is not possible, you may also assign a unique routerid to your separate SUSE OpenStack Cloud 8 system by changing the keepalived_vrrp_offset service configurable. The routerid is currently derived using the keepalived_vrrp_index which comes from a configuration processor variable and the keepalived_vrrp_offset.

For example,

  1. Log in to your Cloud Lifecycle Manager.

  2. Edit your ~/openstack/my_cloud/config/keepalived/defaults.yml file and change the value of the following line:

    keepalived_vrrp_offset: 0

    Change the off value to a number that uniquely identifies a separate vrrp cluster. For example:

    keepalived_vrrp_offset: 0 for the 1st vrrp cluster on this subnet.

    keepalived_vrrp_offset: 1 for the 2nd vrrp cluster on this subnet.

    keepalived_vrrp_offset: 2 for the 3rd vrrp cluster on this subnet.

  3. Commit your configuration to the Git repository (see Book “Installing with Cloud Lifecycle Manager”, Chapter 10 “Using Git for Configuration Management”), as follows:

    ardana > cd ~/openstack/ardana/ansible
    ardana > git add -A
    ardana > git commit -m "changing Admin password"
  4. Run the configuration processor with this command:

    ardana > cd ~/openstack/ardana/ansible
    ardana > ansible-playbook -i hosts/localhost config-processor-run.yml
  5. Use the playbook below to create a deployment directory:

    ardana > cd ~/openstack/ardana/ansible
    ardana > ansible-playbook -i hosts/localhost ready-deployment.yml
  6. If you are making this change after your initial install, run the following reconfigure playbook to make this change in your environment:

    ardana > cd ~/scratch/ansible/next/ardana/ansible/
    ardana > ansible-playbook -i hosts/verb_hosts FND-CLU-reconfigure.yml
Print this page