Distributed Compute Node deployment

Introduction

Additional groups of compute nodes can be deployed and integrated with an existing deployment of a control plane stack. These compute nodes are deployed in separate stacks from the main control plane (overcloud) stack, and they consume exported data from the overcloud stack to reuse as configuration data.

Deploying these additional nodes in separate stacks provides for separation of management between the control plane stack and the stacks for additional compute nodes. The stacks can be managed, scaled, and updated separately.

Using separate stacks also creates smaller failure domains as there are less baremetal nodes in each individual stack. A failure in one baremetal node only requires that management operations to address that failure need only affect the single stack that contains the failed node.

A routed spine and leaf networking layout can be used to deploy these additional groups of compute nodes in a distributed nature. Not all nodes need to be co-located at the same physical location or datacenter. See Deploying Overcloud with L3 routed networking for more details.

Such an architecture is referred to as “Distributed Compute Node” or “DCN” for short.

Supported failure modes and High Availability recommendations

Handling negative scenarios for DCN starts from the deployment planning, like choosing some particular SDN solution over provider networks to meet the expected SLA.

Loss of control plane connectivity

A failure of the central control plane affects all DCN edge sites. There is no autonomous control planes at the edge. No OpenStack control plane API or CLI operations can be executed locally in that case. For example, you cannot create a snapshot of a Nova VM, or issue an auth token, nor can you delete an image or a VM.

Note

A single Controller service failure normally induces no downtime for edge sites and should be handled as for usual HA deployments.

Loss of an edge site

Running Nova VM instances will keep running. If stopped running, you need the control plane back to recover the stopped or crashed workloads. If Neutron DHCP agent is centralized, and we are forwarding DHCP requests to the central site, any VMs that are trying to renew their IPs will eventually time out and lose connectivity.

Note

A single Compute service failure normally affects only its edge site without additional downtime induced for neighbor edge sites or the central control plane.

OpenStack infrastructure services, like Nova Compute, will automatically reconnect to MariaDB database cluster and RabbitMQ broker when the control plane’s uplink is back. No timed out operations can be resumed though and need to be retried manually.

It is recommended to maintain each DCN edge site as a separate Availability Zone (AZ) for Nova/Neutron and Cinder services.

Improving resiliency for N/S and E/W traffic

Reliability of the central control plane may be enhanced with L3 HA network, which only provides North-South routing. The East-West routing effectiveness of edge networks may be improved by using DVR or highly available Open Virtual Network (OVN). There is also BGPVPN and its backend specific choices.

Network recommendations

Traditional or external provider networks with backbone routing at the edge may fulfill or complement a custom distributed routing solution, like L3 Spine-Leaf topology.

Note

Neutron SDN backends that involve tunnelling may be sub-optimal for Edge DCN cases because of the known issues 1808594 and 1808062.

For dynamic IPv4 and stateful IPv6 IPAM cases, you will also need DHCP on those provider networks in order to assign IPs to VM instances. External provider networks usually require no Neutron DHCP agents and handle IPAM (and routing) on its own. While for traditional or Routed Provider Networks, when there is no L2 connectivity to edge over WAN, and Neutron DHCP agents are placed on controllers at the central site, you should have a DHCP relay on every provider network. Alternatively, DHCP agents need to be moved to the edge. Such setups also require highly reliable links between remote and central sites.

Note

Neither of DHCP relays/agents at compute nodes, nor routed/external provider networks are tested or automated via TripleO Heat Templates. You would have to have those configured manually for your DCN environments.

Note

OVN leverages DVR and does not require running Neutron DHCP/L3 agents, which might as well simplify particular DCN setups.

That said, when there is a network failure that disconnects the edge off the central site, there is no SLA for recovery time but only what the provider networks or a particular SDN choice can guarantee. For switched/routed/MPLS provider networks, that may span from 10’s of ms to a few seconds. With the outage thresholds are typically considered to be a 15 seconds. These trace back on various standards that are relevant here.

Config-drive/cloud-init details

Config-drive uses virtual media capabilities of the BMC controller, so that no DHCP is required for VMs to obtain IP addresses at edge sites. This is the most straightforward solution. This does require that the WAN between the remote site and central site is live during deployment of a VM, but after that the VM can run independently without a connection to the central site.

Note

Config-drive may be a tricky for VMs that do not support cloud-init, like some appliance VMs. It may be that such ones (or other VMs that do not support config-drive) will have to be configured with a static IP that matches the Neutron port.

The simplest solution we recommend for DCN would involve only external provider networks at the edge. For that case, it is also recommended to use either config-drive, or IPv6 SLAAC, or another configuration mechanism other than those requiring a 169.254.169.254/32 route for the provider routers to forward data to the metadata service.

IPv6 details

IPv6 for tenants’ workloads and infrastructure tunnels interconnecting the central site and the edge is a viable option as well. IPv6 cannot be used for provisioning networks though. Key benefits IPv6 may provide for DCN are:

• SLAAC, which is a EUI-64 form of autoconfig that makes IPv6 addresses calculated based on MAC addresses and requires no DHCP services placed on the provider networks.

• Improved mobility for endpoints, like NFV APIs, to roam around different links and edge sites without losing its connections and IP addresses.

• End-to-end IPv6 has been shown to have better performance by large content networks. This is largely due to the presence of NAT in most end-to-end IPv4 connections that slows them down.

Storage recommendations

Prior to Ussuri, DCN was only available with ephemeral storage for Nova Compute services. Enhanced data availability, locality awareness and/or replication mechanisms had to be addressed only on the edge cloud application layer.

In Ussuri and newer, TripleO is able to deploy Distributed Multibackend Storage which may be combined with the example in this document to add distributed image management and persistent storage at the edge.

Deploying DCN

Deploying the DCN architecture requires consideration as it relates to the undercloud, roles, networks, and availability zones configuration. This section will document on how to approach the DCN deployment.

The deployment will make use of specific undercloud configuration, and then deploying multiple stacks, typically one stack per distributed location, although this is not a strict requirement.

At the central site, stack separation can still be used to deploy separate stacks for control plane and compute services if compute services are desired at the central site. See deploy_control_plane for more information.

Each distributed site will be a separate stack as well. See deploy_dcn for more information.

Undercloud configuration

This section describes the steps required to configure the undercloud for DCN.

Using direct deploy instead of iSCSI

In a default undercloud configuration, ironic deploys nodes using the iscsi deploy interface. When using the iscsi deploy interface, the deploy ramdisk publishes the node’s disk as an iSCSI target, and the ironic-conductor service then copies the image to this target.

For a DCN deployment, network latency is often a concern between the undercloud and the distributed compute nodes. Considering the potential for latency, the distributed compute nodes should be configured to use the direct deploy interface in the undercloud. This process is described later in this guide under Configure nodes to use the deploy interface.

When using the direct deploy interface, the deploy ramdisk will download the image over HTTP from the undercloud’s Swift service, and copy it to the node’s disk. HTTP is more resilient when dealing with network latency than iSCSI, so using the direct deploy interface provides a more stable node deployment experience for distributed compute nodes.

Configure the Swift temporary URL key

Images used for overcloud deployment are served by Swift and are made available to nodes using an HTTP URL, over the direct deploy interface. To allow Swift to create temporary URLs, it must be configured with a temporary URL key. The key value is used for cryptographic signing and verification of the temporary URLs created by Swift.

The following commands demonstrate how to configure the setting. In this example, uuidgen is used to randomly create a key value. You should choose a unique key value that is a difficult to guess value. For example:

source ~/stackrc
openstack role add --user admin --project service ResellerAdmin
openstack --os-project-name service object store account set --property Temp-URL-Key=$(uuidgen | sha1sum | awk '{​print $1}')

Configure nodes to use the deploy interface

This section describes how to configure the deploy interface for new and existing nodes.

For new nodes, the deploy interface can be specified directly in the JSON structure for each node. For example, see the “deploy_interface”: “direct” setting below:

{
   "nodes":[
       {
           "mac":[
               "bb:bb:bb:bb:bb:bb"
           ],
           "name":"node01",
           "cpu":"4",
           "memory":"6144",
           "disk":"40",
           "arch":"x86_64",
           "pm_type":"ipmi",
           "pm_user":"admin",
           "pm_password":"p@55w0rd!",
           "pm_addr":"192.168.24.205",
           “deploy_interface”: “direct”
       },
       {
           "mac":[
               "cc:cc:cc:cc:cc:cc"
           ],
           "name":"node02",
           "cpu":"4",
           "memory":"6144",
           "disk":"40",
           "arch":"x86_64",
           "pm_type":"ipmi",
           "pm_user":"admin",
           "pm_password":"p@55w0rd!",
           "pm_addr":"192.168.24.206"
           “deploy_interface”: “direct”
       }
   ]
}

Existing nodes can be updated to use the direct deploy interface. For example:

baremetal node set --deploy-interface direct 4b64a750-afe3-4236-88d1-7bb88c962666

Deploying the control plane

The main overcloud control plane stack should be deployed as needed for the desired cloud architecture layout. This stack contains nodes running the control plane and infrastructure services needed for the cloud. For the purposes of this documentation, this stack is referred to as the control-plane stack.

No specific changes or deployment configuration is necessary to deploy just the control plane services.

It’s possible to configure the control-plane stack to contain only control plane services, and no compute or storage services. If compute and storage services are desired at the same geographical site as the control-plane stack, then they may be deployed in a separate stack just like a edge site specific stack, but using nodes at the same geographical location. In such a scenario, the stack with compute and storage services could be called central and deploying it in a separate stack allows for separation of management and operations. This scenario may also be implemented with an “external” Ceph cluster for storage as described in Use an external Ceph cluster with the Overcloud. If however, Glance needs to be configured with multiple stores so that images may be served to remote sites one control-plane stack may be used as described in Distributed Multibackend Storage.

It is suggested to give each stack an explicit name. For example, the control plane stack could be called control-plane and set by passing --stack control-plane to the openstack overcloud deploy command.

Deploying a DCN site

Once the control plane is deployed, separate deployments can be done for each DCN site. This section will document how to perform such a deployment.

Saving configuration from the overcloud

Once the overcloud control plane has been deployed, data needs to be retrieved from the overcloud Heat stack and plan to pass as input values into the separate DCN deployment.

Beginning in Wallaby with Ephemeral Heat, the export file is created automatically under the default working directory which defaults to $HOME/overcloud-deploy/<stack>. The working directory can also be set with the --working-dir cli argument to the openstack overcloud deploy command.

The export file will be automatically created as $HOME/overcloud-deploy/<stack>/<stack>-export.yaml.

Victoria and prior releases

In Victoria and prior releases, the export file must be created by running the export command.

Extract the needed data from the control plane stack:

# Pass --help to see a full list of options
openstack overcloud export \
  --stack control-plane \
  --output-file control-plane-export.yaml

Note

The generated control-plane-export.yaml contains sensitive security data such as passwords and TLS certificates that are used in the overcloud deployment. Some passwords in the file may be removed if they are not needed by DCN. For example, the passwords for RabbitMQ, MySQL, Keystone, Nova and Neutron should be sufficient to launch an instance. When the export common is run, the Ceph passwords are excluded so that DCN deployments which include Ceph do not reuse the same Ceph password and instead new ones are generated per DCN deployment.

Care should be taken to keep the file as secured as possible.

Network resource configuration

Beginning in Wallaby, the Network V2 model is used to provision and manage the network related resources (networks, subnets, and ports). Specifying a parameter value for ManageNetworks or using the external resource UUID’s is deprecated in Wallaby.

The network resources can either be provisioned and managed with the separate openstack overcloud network command or as part of the openstack overcloud deploy command using the cli args (--networks-file, --vip-file, --baremetal-deployment).

See Networking Version 2 (Two) for the full details on managing the network resources.

With Network v2, the Heat stack no longer manages any of the network resources. As such with using DCN with multi-stack, it is no longer necessary to first update the central stack to provision new network resources when deploying a new site. Instead, it is all handled as part of the network provisioning commands or the overcloud deploy command.

The same files used with the cli args (--networks-file, --vip-file, --baremetal-deployment), should be the same files used across all stacks in a DCN deployment.

Victoria and prior releases

When deploying separate stacks it may be necessary to reuse networks, subnets, and VIP resources between stacks if desired. Only a single Heat stack can own a resource and be responsible for its creation and deletion, however the resources can be reused in other stacks.

ManageNetworks

The ManageNetworks parameter can be set to false so that the same network_data.yaml file can be used across all the stacks. When ManageNetworks is set to false, ports will be created for the nodes in the separate stacks on the existing networks that were already created in the control-plane stack.

When ManageNetworks is used, it’s a global option for the whole stack and applies to all of the network, subnet, and segment resources.

To use ManageNetworks, create an environment file which sets the parameter value to false:

parameter_defaults:
  ManageNetworks: false

When using ManageNetworks, all network resources (except for ports) are managed in the central stack. When the central stack is deployed, ManageNetworks should be left unset (or set to True). When a child stack is deployed, it is then set to false so that the child stack does not attempt to manage the already existing network resources.

Additionally, when adding new network resources, such as entire new leaves when deploying spine/leaf, the central stack must first be updated with the new network_data.yaml that contains the new leaf definitions. Even though the central stack is not directly using the new network resources, it still is responsible for creating and managing them. Once the new network resources are made available in the central stack, a child stack (such as a new edge site) could be deployed using the new networks.

External UUID’s

If more fine grained control over which networks should be reused from the control-plane stack is needed, then various external_resource_* fields can be added to network_data.yaml. When these fields are present on network, subnet, segment, or vip resources, Heat will mark the resources in the separate stack as being externally managed, and it won’t try to any create, update, or delete operations on those resources.

ManageNetworks should not be set when when the external_resource_* fields are used.

The external resource fields that can be used in network_data.yaml are as follows:

external_resource_network_id: Existing Network UUID
external_resource_subnet_id: Existing Subnet UUID
external_resource_segment_id: Existing Segment UUID
external_resource_vip_id: Existing VIP UUID

These fields can be set on each network definition in the network_data.yaml` file used for the deployment of the separate stack.

Not all networks need to be reused or shared across stacks. The external_resource_* fields can be set for only the networks that are meant to be shared, while the other networks can be newly created and managed.

For example, to reuse the internal_api network from the control plane stack in a separate stack, run the following commands to show the UUIDs for the related network resources:

openstack network show internal_api -c id -f value
openstack subnet show internal_api_subnet -c id -f value
openstack port show internal_api_virtual_ip -c id -f value

Save the values shown in the output of the above commands and add them to the network definition for the internal_api network in the network_data.yaml file for the separate stack. An example network definition would look like:

- name: InternalApi
  external_resource_network_id: 93861871-7814-4dbc-9e6c-7f51496b43af
  external_resource_subnet_id: c85c8670-51c1-4b17-a580-1cfb4344de27
  external_resource_vip_id: 8bb9d96f-72bf-4964-a05c-5d3fed203eb7
  name_lower: internal_api
  vip: true
  ip_subnet: '172.16.2.0/24'
  allocation_pools: [{'start': '172.16.2.4', 'end': '172.16.2.250'}]
  ipv6_subnet: 'fd00:fd00:fd00:2000::/64'
  ipv6_allocation_pools: [{'start': 'fd00:fd00:fd00:2000::10', 'end': 'fd00:fd00:fd00:2000:ffff:ffff:ffff:fffe'}]
  mtu: 1400

Note

When not sharing networks between stacks, each network defined in network_data.yaml must have a unique name across all deployed stacks. This requirement is necessary since regardless of the stack, all networks are created in the same tenant in Neutron on the undercloud.

For example, the network name internal_api can’t be reused between stacks, unless the intent is to share the network between the stacks. The network would need to be given a different name and name_lower property such as InternalApiCompute0 and internal_api_compute_0.

If separate storage and storage management networks are used with multiple Ceph clusters and Glance servers per site, then a routed storage network should be shared between sites for image transfer. The storage management network, which Ceph uses to keep OSDs balanced, does not need to be shared between sites.

DCN related roles

Different roles are provided within tripleo-heat-templates, depending on the configuration and desired services to be deployed at each distributed site.

The default compute role at roles/Compute.yaml can be used if that is sufficient for the use case.

Three additional roles are also available for deploying compute nodes with co-located persistent storage at the distributed site.

The first is roles/DistributedCompute.yaml. This role includes the default compute services, but also includes the cinder volume service. The cinder volume service would be configured to talk to storage that is local to the distributed site for persistent storage.

The second is roles/DistributedComputeHCI.yaml. This role includes the default computes services, the cinder volume service, and also includes the Ceph Mon, Mgr, and OSD services for deploying a Ceph cluster at the distributed site. Using this role, both the compute services and Ceph services are deployed on the same nodes, enabling a hyperconverged infrastructure for persistent storage at the distributed site. When Ceph is used, there must be a minimum of three DistributedComputeHCI nodes. This role also includes a Glance server, provided by the GlanceApiEdge service with in the DistributedComputeHCI role. The Nova compute service of each node in the DistributedComputeHCI role is configured by default to use its local Glance server.

The third is roles/DistributedComputeHCIScaleOut.yaml. This role is like the DistributedComputeHCI role but does not run the Ceph Mon and Mgr service. It offers the Ceph OSD service however, so it may be used to scale up storage and compute services at each DCN site after the minimum of three DistributedComputeHCI nodes have been deployed. There is no GlanceApiEdge service in the DistributedComputeHCIScaleOut role but in its place the Nova compute service of the role is configured by default to connect to a local HaProxyEdge service which in turn proxies image requests to the Glance servers running on the DistributedComputeHCI roles.

For information on configuring the distributed Glance services see Distributed Multibackend Storage.

Configuring Availability Zones (AZ)

Each edge site must be configured as a separate availability zone (AZ). When you deploy instances to this AZ, you can expect it to run on the remote Compute node. In addition, the central site must also be within a specific AZ (or multiple AZs), rather than the default AZ.

When also deploying persistent storage at each site, the storage backend availability zone must match the compute availability zone name.

AZs are configured differently for compute (Nova) and storage (Cinder). Configuring AZs are documented in the next sections.

Configuring AZs for Nova (compute)

The Nova AZ configuration for compute nodes in the stack can be set with the NovaComputeAvailabilityZone parameter during the deployment.

The value of the parameter is the name of the AZ where compute nodes in that stack will be added.

For example, the following environment file would be used to add compute nodes in the edge-0 stack to the edge-0 AZ:

parameter_defaults:
   NovaComputeAvailabilityZone: edge-0

Additionally, the OS::TripleO::NovaAZConfig service must be enabled by including the following resource_registry mapping:

resource_registry:
  OS::TripleO::Services::NovaAZConfig: tripleo-heat-templates/deployment/nova/nova-az-config.yaml

Or, the following environment can be included which sets the above mapping:

environments/nova-az-config.yaml

It’s also possible to configure the AZ for a compute node by adding it to a host aggregate after the deployment is completed. The following commands show creating a host aggregate, an associated AZ, and adding compute nodes to a edge-0 AZ:

openstack aggregate create edge-0 --zone edge-0
openstack aggregate add host edge-0 hostA
openstack aggregate add host edge-0 hostB

Note

The above commands are run against the deployed overcloud, not the undercloud. Make sure the correct rc file for the control plane stack of the overcloud is sourced for the shell before running the commands.

Configuring AZs for Cinder (storage)

Each site that uses consistent storage is configured with its own cinder backend(s). Cinder backends are not shared between sites. Each backend is also configured with an AZ that should match the configured Nova AZ for the compute nodes that will make use of the storage provided by that backend.

The CinderStorageAvailabilityZone parameter can be used to configure the AZ for a given backend. Parameters are also available for different backend types, such as CinderISCSIAvailabilityZone, CinderRbdAvailabilityZone, and CinderNfsAvailabilityZone. When set, the backend type specific parameter will take precedence over CinderStorageAvailabilityZone.

This example shows an environment file setting the AZ for the backend in the central site:

parameter_defaults:
   CinderStorageAvailabilityZone: central

This example shows an environment file setting the AZ for the backend in the edge0 site:

parameter_defaults:
   CinderStorageAvailabilityZone: edge0

Deploying Ceph with HCI

When deploying Ceph while using the DistributedComputeHCI and DistributedComputeHCIScaleOut roles, the following environment file should be used to enable Ceph:

environments/ceph-ansible/ceph-ansible.yaml

Sample environments

There are sample environments that are included in tripleo-heat-templates for setting many of the parameter values and resource_registry mappings. These environments are located within the tripleo-heat-templates directory at:

environments/dcn.yaml
environments/dcn-storage.yaml

The environments are not all-inclusive and do not set all needed values and mappings, but can be used as a guide when deploying DCN.

Example: DCN deployment with pre-provisioned nodes, shared networks, and multiple stacks

This example shows the deployment commands and associated environment files of a example of a DCN deployment. The deployment uses pre-provisioned nodes. All networks are shared between the multiple stacks. The example illustrates the deployment workflow of deploying multiple stacks for a real world DCN deployment.

Four stacks are deployed:

control-plane

All control plane services. Shares the same geographical location as the central stack.

central

Compute, Cinder, Ceph deployment. Shares the same geographical location as the control-plane stack.

edge0

Compute, Cinder, Ceph deployment. Separate geographical location from any other stack.

edge1

Compute, Cinder, Ceph deployment. Separate geographical location from any other stack.

Notice how the central stack will contain only compute and storage services. It is really just another instance of an edge site, but just happens to be deployed at the same geographical location as the control-plane stack. control-plane and central could instead be deployed in the same stack, however for easier manageability and separation, they are deployed in separate stacks.

This example also uses pre-provisioned nodes as documented at Using Already Deployed Servers.

Undercloud

Since this example uses pre-provisioned nodes, no additional undercloud configuration is needed. The steps in undercloud_dcn are not specifically applicable when using pre-provisioned nodes.

Deploy the control-plane stack

The control-plane stack is deployed with the following command:

openstack overcloud deploy \
  --verbose \
  --stack control-plane \
  --disable-validations \
  --templates /home/centos/tripleo-heat-templates \
  -r roles-data.yaml \
  -e role-counts.yaml \
  --networks-file network_data_v2.yaml \
  --vip-file vip_data.yaml \
  --baremetal-deployment baremetal_deployment.yaml \
  -e /home/centos/tripleo-heat-templates/environments/docker-ha.yaml \
  -e /home/centos/tripleo/environments/containers-prepare-parameter.yaml \
  -e /home/centos/tripleo-heat-templates/environments/deployed-server-environment.yaml \
  -e /home/centos/tripleo-heat-templates/environments/deployed-server-bootstrap-environment-centos.yaml \
  -e /home/centos/tripleo-heat-templates/environments/network-isolation.yaml \
  -e /home/centos/tripleo-heat-templates/environments/net-multiple-nics.yaml \
  -e hostnamemap.yaml \
  -e network-environment.yaml \
  -e deployed-server-port-map.yaml \
  -e az.yaml

Many of the specified environments and options are not specific to DCN. The ones that relate to DCN are as follows.

--stack control-plane sets the stack name to control-plane.

The roles-data.yaml file contains only the Controller role from the templates directory at roles/Controller.yaml.

role-counts.yaml contains:

parameter_defaults:
  ControllerCount: 1

Warning

Only one Controller node is deployed for example purposes but three are recommended in order to have a highly available control plane.

network_data_v2.yaml, vip_data.yaml, and baremetal_deployment.yaml contain the definitions to manage the network resources. See Networking Version 2 (Two) for creating these files.

az.yaml contains:

parameter_defaults:
  CinderStorageAvailabilityZone: 'central'
  NovaComputeAvailabilityZone: 'central'

When the deployment completes, a single stack is deployed:

(undercloud) [centos@scale ~]$ openstack stack list
+--------------------------------------+---------------+----------------------------------+-----------------+----------------------+----------------------+
| ID                                   | Stack Name    | Project | Stack Status    | Creation Time        | Updated Time         |
+--------------------------------------+---------------+----------------------------------+-----------------+----------------------+----------------------+
| 5f172fd8-97a5-4b9b-8d4c-2c931fd048e7 | control-plane | c117a9b489384603b2f45185215e9728 | CREATE_COMPLETE | 2019-03-13T18:51:08Z | 2019-03-13T19:44:27Z |
+--------------------------------------+---------------+----------------------------------+-----------------+----------------------+----------------------+

Exported configuration from the control-plane stack

As documented in export_dcn, the export file is created automatically by the openstack overcloud deploy command.

Victoria and prior releases

For Victoria and prior releases, the following command must be run to export the configuration data from the control-plane stack:

openstack overcloud export \
  --stack control-plane \
  --output-file control-plane-export.yaml

Deploy the central stack

The central stack deploys compute and storage services to be co-located at the same site where the control-plane stack was deployed.

The central stack is deployed with the following command:

openstack overcloud deploy \
  --verbose \
  --stack central \
  --templates /home/centos/tripleo-heat-templates \
  -r distributed-roles-data.yaml \
  -n site_network_data.yaml \
  --disable-validations \
  --networks-file network_data_v2.yaml \
  --vip-file vip_data.yaml \
  --baremetal-deployment baremetal_deployment.yaml \
  -e /home/centos/tripleo-heat-templates/environments/docker-ha.yaml \
  -e /home/centos/tripleo/environments/containers-prepare-parameter.yaml \
  -e /home/centos/tripleo-heat-templates/environments/deployed-server-environment.yaml \
  -e /home/centos/tripleo-heat-templates/environments/deployed-server-bootstrap-environment-centos.yaml \
  -e /home/centos/tripleo-heat-templates/environments/network-isolation.yaml \
  -e /home/centos/tripleo-heat-templates/environments/net-multiple-nics.yaml \
  -e /home/centos/tripleo-heat-templates/environments/ceph-ansible/ceph-ansible.yaml \
  -e /home/centos/tripleo-heat-templates/environments/low-memory-usage.yaml \
  -e role-counts.yaml \
  -e hostnamemap.yaml \
  -e network-environment.yaml \
  -e deployed-server-port-map.yaml \
  -e ceph-environment.yaml \
  -e az.yaml \
  -e /home/centos/overcloud-deploy/control-plane/control-plane-export.yaml

--stack central sets the stack name to central.

distributed-roles-data.yaml contains a single role called DistributedComputeHCI which contains Nova, Cinder, and Ceph services. The example role is from the templates directory at roles/DistributedComputeHCI.yaml.

role-counts.yaml contains:

parameter_defaults:
  DistributedComputeHCICount: 1

Warning

Only one DistributedComputeHCI is deployed for example purposes but three are recommended in order to have a highly available Ceph cluster. If more than three such nodes of that role are necessary for additional compute and storage resources, then use additional nodes from the DistributedComputeHCIScaleOut role.

network_data_v2.yaml, vip_data.yaml, and baremetal_deployment.yaml are the same files used with the control-plane stack.

az.yaml contains the same content as was used in the control-plane stack:

parameter_defaults:
  CinderStorageAvailabilityZone: 'central'
  NovaComputeAvailabilityZone: 'central'

The control-plane-export.yaml file was generated by the openstack overcloud deploy command when deploying the control-plane stack or from the command from example_export_dcn (Victoria or prior releases).

When the deployment completes, 2 stacks are deployed:

+--------------------------------------+---------------+----------------------------------+-----------------+----------------------+----------------------+
| ID                                   | Stack Name    | Project                          | Stack Status    | Creation Time        | Updated Time         |
+--------------------------------------+---------------+----------------------------------+-----------------+----------------------+----------------------+
| 0bdade63-4645-4490-a540-24be48527e10 | central       | c117a9b489384603b2f45185215e9728 | CREATE_COMPLETE | 2019-03-25T21:35:49Z | None                 |
| 5f172fd8-97a5-4b9b-8d4c-2c931fd048e7 | control-plane | c117a9b489384603b2f45185215e9728 | CREATE_COMPLETE | 2019-03-13T18:51:08Z | None                 |
+--------------------------------------+---------------+----------------------------------+-----------------+----------------------+----------------------+

The AZ and aggregate configuration for Nova can be checked and verified with these commands. Note that the rc file for the control-plane stack must be sourced as these commands talk to overcloud APIs:

(undercloud) [centos@scale ~]$ source control-planerc
(control-plane) [centos@scale ~]$ openstack aggregate list
+----+---------+-------------------+
| ID | Name    | Availability Zone |
+----+---------+-------------------+
|  9 | central | central           |
+----+---------+-------------------+
(control-plane) [centos@scale ~]$ openstack aggregate show central
+-------------------+----------------------------+
| Field             | Value                      |
+-------------------+----------------------------+
| availability_zone | central                    |
| created_at        | 2019-03-25T22:23:25.000000 |
| deleted           | False                      |
| deleted_at        | None                       |
| hosts             | [u'compute-0.localdomain'] |
| id                | 9                          |
| name              | central                    |
| properties        |                            |
| updated_at        | None                       |
+-------------------+----------------------------+
(control-plane) [centos@scale ~]$ nova availability-zone-list
+----------------------------+----------------------------------------+
| Name                       | Status                                 |
+----------------------------+----------------------------------------+
| internal                   | available                              |
| |- openstack-0.localdomain |                                        |
| | |- nova-conductor        | enabled :-) 2019-03-27T18:21:29.000000 |
| | |- nova-scheduler        | enabled :-) 2019-03-27T18:21:31.000000 |
| | |- nova-consoleauth      | enabled :-) 2019-03-27T18:21:34.000000 |
| central                    | available                              |
| |- compute-0.localdomain   |                                        |
| | |- nova-compute          | enabled :-) 2019-03-27T18:21:32.000000 |
+----------------------------+----------------------------------------+
(control-plane) [centos@scale ~]$ openstack compute service list
+----+------------------+-------------------------+----------+---------+-------+----------------------------+
| ID | Binary           | Host                    | Zone     | Status  | State | Updated At                 |
+----+------------------+-------------------------+----------+---------+-------+----------------------------+
|  1 | nova-scheduler   | openstack-0.localdomain | internal | enabled | up    | 2019-03-27T18:23:31.000000 |
|  2 | nova-consoleauth | openstack-0.localdomain | internal | enabled | up    | 2019-03-27T18:23:34.000000 |
|  3 | nova-conductor   | openstack-0.localdomain | internal | enabled | up    | 2019-03-27T18:23:29.000000 |
|  7 | nova-compute     | compute-0.localdomain   | central  | enabled | up    | 2019-03-27T18:23:32.000000 |
+----+------------------+-------------------------+----------+---------+-------+----------------------------+

Note how a new aggregate and AZ called central has been automatically created. The newly deployed nova-compute service from the compute-0 host in the central stack is automatically added to this aggregate and zone.

The AZ configuration for Cinder can be checked and verified with these commands:

(control-plane) [centos@scale ~]$ openstack volume service list
+------------------+-------------------------+---------+---------+-------+----------------------------+
| Binary           | Host                    | Zone    | Status  | State | Updated At                 |
+------------------+-------------------------+---------+---------+-------+----------------------------+
| cinder-scheduler | openstack-0.rdocloud    | central | enabled | up    | 2019-03-27T21:17:44.000000 |
| cinder-volume    | compute-0@tripleo_ceph  | central | enabled | up    | 2019-03-27T21:17:44.000000 |
+------------------+-------------------------+---------+---------+-------+----------------------------+

The Cinder AZ configuration shows the ceph backend in the central zone which was deployed by the central stack.

Deploy the edge-0 and edge-1 stacks

Now that the control-plane and central stacks are deployed, we’ll deploy an edge-0 and edge-1 stack. These stacks are similar to the central stack in that they deploy the same roles with the same services. It differs in that the nodes will be managed in a separate stack and it illustrates the separation of deployment and management between edge sites.

The AZs will be configured differently in these stacks as the nova and cinder services will belong to an AZ unique to each the site.

The edge-0 stack is deployed with the following command:

openstack overcloud deploy \
  --verbose \
  --stack edge-0 \
  --templates /home/centos/tripleo-heat-templates \
  -r distributed-roles-data.yaml \
  -n site_network_data.yaml \
  --disable-validations \
  --networks-file network_data_v2.yaml \
  --vip-file vip_data.yaml \
  --baremetal-deployment baremetal_deployment.yaml \
  -e /home/centos/tripleo-heat-templates/environments/docker-ha.yaml \
  -e /home/centos/tripleo/environments/containers-prepare-parameter.yaml \
  -e /home/centos/tripleo-heat-templates/environments/deployed-server-environment.yaml \
  -e /home/centos/tripleo-heat-templates/environments/deployed-server-bootstrap-environment-centos.yaml \
  -e /home/centos/tripleo-heat-templates/environments/network-isolation.yaml \
  -e /home/centos/tripleo-heat-templates/environments/net-multiple-nics.yaml \
  -e /home/centos/tripleo-heat-templates/environments/ceph-ansible/ceph-ansible.yaml \
  -e /home/centos/tripleo-heat-templates/environments/low-memory-usage.yaml \
  -e role-counts.yaml \
  -e hostnamemap.yaml \
  -e network-environment.yaml \
  -e deployed-server-port-map.yaml \
  -e ceph-environment.yaml \
  -e az.yaml \
  -e /home/centos/overcloud-deploy/control-plane/control-plane-export.yaml

--stack edge-0 sets the stack name to edge-0.

distributed-roles-data.yaml contains a single role called DistributedComputeHCI which contains Nova, Cinder, and Ceph services. The example role is from the templates directory at roles/DistributedComputeHCI.yaml. This file is the same as was used in the central stack.

role-counts.yaml contains:

parameter_defaults:
  DistributedComputeHCICount: 1

Warning

Only one DistributedComputeHCI is deployed for example purposes but three are recommended in order to have a highly available Ceph cluster. If more than three such nodes of that role are necessary for additional compute and storage resources, then use additional nodes from the DistributedComputeHCIScaleOut role.

az.yaml contains specific content for the edge-0 stack:

parameter_defaults:
  CinderStorageAvailabilityZone: 'edge-0'
  NovaComputeAvailabilityZone: 'edge-0'

The CinderStorageAvailabilityZone and NovaDefaultAvailabilityZone parameters are set to edge-0 to match the stack name.

The control-plane-export.yaml file was generated by the openstack overcloud deploy command when deploying the control-plane stack or from the command from example_export_dcn (Victoria or prior releases).

network_data_v2.yaml, vip_data.yaml, and baremetal_deployment.yaml are the same files used with the control-plane stack. However, the files will need to be modified if the edge-0 or edge-1 stacks require addditional provisioning of network resources for new subnets. Update the files as needed and continue to use the same files for all overcloud deploy commands across all stacks.

The edge-1 stack is deployed with a similar command. The stack is given a different name with --stack edge-1 and az.yaml contains:

parameter_defaults:
  CinderStorageAvailabilityZone: 'edge-1'
  NovaComputeAvailabilityZone: 'edge-1'

When the deployment completes, there are now 4 stacks are deployed:

(undercloud) [centos@scale ~]$ openstack stack list
+--------------------------------------+---------------+----------------------------------+-----------------+----------------------+----------------------+
| ID                                   | Stack Name    | Project                          | Stack Status    | Creation Time        | Updated Time         |
+--------------------------------------+---------------+----------------------------------+-----------------+----------------------+----------------------+
| 203dc480-3b0b-4cd9-9f70-f79898461c17 | edge-0        | c117a9b489384603b2f45185215e9728 | CREATE_COMPLETE | 2019-03-29T17:30:15Z | None                 |
| 0bdade63-4645-4490-a540-24be48527e10 | central       | c117a9b489384603b2f45185215e9728 | CREATE_COMPLETE | 2019-03-25T21:35:49Z | None                 |
| 5f172fd8-97a5-4b9b-8d4c-2c931fd048e7 | control-plane | c117a9b489384603b2f45185215e9728 | UPDATE_COMPLETE | 2019-03-13T18:51:08Z | 2019-03-13T19:44:27Z |
+--------------------------------------+---------------+----------------------------------+-----------------+----------------------+----------------------+

Repeating the same commands that were run after the central stack was deployed to show the AZ configuration, shows that the new AZs for edge-0 and edge-1 are created and available:

(undercloud) [centos@scale ~]$ source control-planerc
(control-plane) [centos@scale ~]$ openstack aggregate list
+----+---------+-------------------+
| ID | Name    | Availability Zone |
+----+---------+-------------------+
|  9 | central | central           |
| 10 | edge-0  | edge-0            |
| 11 | edge-1  | edge-1            |
+----+---------+-------------------+
(control-plane) [centos@scale ~]$ openstack aggregate show edge-0
+-------------------+----------------------------+
| Field             | Value                      |
+-------------------+----------------------------+
| availability_zone | edge-0                     |
| created_at        | 2019-03-29T19:01:59.000000 |
| deleted           | False                      |
| deleted_at        | None                       |
| hosts             | [u'compute-1.localdomain'] |
| id                | 10                         |
| name              | edge-0                     |
| properties        |                            |
| updated_at        | None                       |
+-------------------+----------------------------+
(control-plane) [centos@scale ~]$ nova availability-zone-list
+----------------------------+----------------------------------------+
| Name                       | Status                                 |
+----------------------------+----------------------------------------+
| internal                   | available                              |
| |- openstack-0.localdomain |                                        |
| | |- nova-conductor        | enabled :-) 2019-04-01T17:38:06.000000 |
| | |- nova-scheduler        | enabled :-) 2019-04-01T17:38:13.000000 |
| | |- nova-consoleauth      | enabled :-) 2019-04-01T17:38:09.000000 |
| central                    | available                              |
| |- compute-0.localdomain   |                                        |
| | |- nova-compute          | enabled :-) 2019-04-01T17:38:07.000000 |
| edge-0                     | available                              |
| |- compute-1.localdomain   |                                        |
| | |- nova-compute          | enabled :-) 2019-04-01T17:38:07.000000 |
| edge-1                     | available                              |
| |- compute-2.localdomain   |                                        |
| | |- nova-compute          | enabled :-) 2019-04-01T17:38:06.000000 |
+----------------------------+----------------------------------------+
(control-plane) [centos@scale ~]$ openstack compute service list
+----+------------------+-------------------------+----------+---------+-------+----------------------------+
| ID | Binary           | Host                    | Zone     | Status  | State | Updated At                 |
+----+------------------+-------------------------+----------+---------+-------+----------------------------+
|  1 | nova-scheduler   | openstack-0.localdomain | internal | enabled | up    | 2019-04-01T17:38:23.000000 |
|  2 | nova-consoleauth | openstack-0.localdomain | internal | enabled | up    | 2019-04-01T17:38:19.000000 |
|  3 | nova-conductor   | openstack-0.localdomain | internal | enabled | up    | 2019-04-01T17:38:26.000000 |
|  7 | nova-compute     | compute-0.localdomain   | central  | enabled | up    | 2019-04-01T17:38:27.000000 |
| 16 | nova-compute     | compute-1.localdomain   | edge-0   | enabled | up    | 2019-04-01T17:38:27.000000 |
| 17 | nova-compute     | compute-2.localdomain   | edge-1   | enabled | up    | 2019-04-01T17:38:26.000000 |
+----+------------------+-------------------------+----------+---------+-------+----------------------------+
(control-plane) [centos@scale ~]$ openstack volume service list
+------------------+-------------------------+---------+---------+-------+----------------------------+
| Binary           | Host                    | Zone    | Status  | State | Updated At                 |
+------------------+-------------------------+---------+---------+-------+----------------------------+
| cinder-scheduler | openstack-0.rdocloud    | central | enabled | up    | 2019-04-01T17:38:27.000000 |
| cinder-volume    | hostgroup@tripleo_iscsi | central | enabled | up    | 2019-04-01T17:38:27.000000 |
| cinder-volume    | compute-0@tripleo_ceph  | central | enabled | up    | 2019-04-01T17:38:30.000000 |
| cinder-volume    | compute-1@tripleo_ceph  | edge-0  | enabled | up    | 2019-04-01T17:38:32.000000 |
| cinder-volume    | compute-2@tripleo_ceph  | edge-1  | enabled | up    | 2019-04-01T17:38:28.000000 |
+------------------+-------------------------+---------+---------+-------+----------------------------+
(control-plane) [centos@scale ~]$

For information on extending this example with distributed image management for image sharing between DCN site Ceph clusters see Distributed Multibackend Storage.

Updating DCN

Each stack in a multi-stack DCN deployment must be updated to perform a full minor update across the entire deployment.

The minor update procedure as detailed in Updating Content on Overcloud Nodes be run for each stack in the deployment.

The control-plane or central stack should be updated first by completing all the steps from the minor update procedure.

Victoria and prior releases

Once the central stack is updated, re-run the export command from Saving configuration from the overcloud to recreate the required input file for each separate DCN stack.

Note

When re-running the export command, save the generated file in a new directory so that the previous version is not overwritten. In the event that a separate DCN stack needs a stack update operation performed prior to the minor update procedure, the previous version of the exported file should be used.

Each separate DCN stack can then be updated individually as required. There is no requirement as to the order of which DCN stack is updated first.

Running Ansible across multiple DCN stacks

Warning

This currently is only supported in Train or newer versions.

Each DCN stack should usually be updated individually. However if you need to run Ansible on nodes deployed from more than one DCN stack, then the tripleo-ansible-inventory command’s --stack option supports being passed more than one stack. If more than one stack is passed, then a single merged inventory will be generated which contains the union of the nodes in those stacks. For example, if you were to run the following:

tripleo-ansible-inventory --static-yaml-inventory inventory.yaml --stack central,edge0

then you could use the generated inventory.yaml as follows:

(undercloud) [stack@undercloud ~]$ ansible -i inventory.yaml -m ping central
central-controller-0 | SUCCESS => {
    "ansible_facts": {
        "discovered_interpreter_python": "/usr/bin/python"
    },
    "changed": false,
    "ping": "pong"
}
(undercloud) [stack@undercloud ~]$ ansible -i inventory.yaml -m ping edge0
edge0-distributedcomputehci-0 | SUCCESS => {
    "ansible_facts": {
        "discovered_interpreter_python": "/usr/bin/python"
    },
    "changed": false,
    "ping": "pong"
}
(undercloud) [stack@undercloud ~]$ ansible -i inventory.yaml -m ping all
undercloud | SUCCESS => {
    "changed": false,
    "ping": "pong"
}
edge0-distributedcomputehci-0 | SUCCESS => {
    "ansible_facts": {
        "discovered_interpreter_python": "/usr/bin/python"
    },
    "changed": false,
    "ping": "pong"
}
central-controller-0 | SUCCESS => {
    "ansible_facts": {
        "discovered_interpreter_python": "/usr/bin/python"
    },
    "changed": false,
    "ping": "pong"
}
(undercloud) [stack@undercloud ~]$

When multiple stacks are passed as input a host group is created after each stack which refers to all of the nodes in that stack. In the example above, edge0 has only one node from the DistributedComputeHci role and central has only one node from the Controller role.

The inventory will also have a host group created for every item in the cross product of stacks and roles. For example, central_Controller, edge0_Compute, edge1_Compute, etc. This is done in order to avoid name collisions, e.g. Compute would refer to all nodes in the Compute role, but when there’s more than one stack edge0_Compute and edge1_Compute refer to different Compute nodes based on the stack from which they were deployed.