c12s-kubespray/docs/openstack.md
Kenichi Omichi 68cfb9a053
Update OpenStack doc for external cloud provider (#6252)
Now the in-tree cloud provider is deprecated and it is recommended to
the external cloud provider for OpenStack instead.
The doc described how to upgrade from the in-tree cloud provider, but
it is better to describe how to deploy the external cloud provider from
scratch instead for current situation.
This updates the OpenStack doc for this usecase.
2020-06-22 04:48:39 -07:00

5.5 KiB

OpenStack

The in-tree cloud provider

To deploy Kubespray on OpenStack uncomment the cloud_provider option in group_vars/all/all.yml and set it to openstack.

After that make sure to source in your OpenStack credentials like you would do when using nova-client or neutron-client by using source path/to/your/openstack-rc or . path/to/your/openstack-rc.

For those who prefer to pass the OpenStack CA certificate as a string, one can base64 encode the cacert file and store it in the variable openstack_cacert.

The next step is to make sure the hostnames in your inventory file are identical to your instance names in OpenStack. Otherwise cinder won't work as expected.

Unless you are using calico or kube-router you can now run the playbook.

Additional step needed when using calico or kube-router:

Being L3 CNI, calico and kube-router do not encapsulate all packages with the hosts' ip addresses. Instead the packets will be routed with the PODs ip addresses directly.

OpenStack will filter and drop all packets from ips it does not know to prevent spoofing.

In order to make L3 CNIs work on OpenStack you will need to tell OpenStack to allow pods packets by allowing the network they use.

First you will need the ids of your OpenStack instances that will run kubernetes:

openstack server list --project YOUR_PROJECT
+--------------------------------------+--------+----------------------------------+--------+-------------+
| ID                                   | Name   | Tenant ID                        | Status | Power State |
+--------------------------------------+--------+----------------------------------+--------+-------------+
| e1f48aad-df96-4bce-bf61-62ae12bf3f95 | k8s-1  | fba478440cb2444a9e5cf03717eb5d6f | ACTIVE | Running     |
| 725cd548-6ea3-426b-baaa-e7306d3c8052 | k8s-2  | fba478440cb2444a9e5cf03717eb5d6f | ACTIVE | Running     |

Then you can use the instance ids to find the connected neutron ports (though they are now configured through using OpenStack):

openstack port list -c id -c device_id --project YOUR_PROJECT
+--------------------------------------+--------------------------------------+
| id                                   | device_id                            |
+--------------------------------------+--------------------------------------+
| 5662a4e0-e646-47f0-bf88-d80fbd2d99ef | e1f48aad-df96-4bce-bf61-62ae12bf3f95 |
| e5ae2045-a1e1-4e99-9aac-4353889449a7 | 725cd548-6ea3-426b-baaa-e7306d3c8052 |

Given the port ids on the left, you can set the two allowed-address(es) in OpenStack. Note that you have to allow both kube_service_addresses (default 10.233.0.0/18) and kube_pods_subnet (default 10.233.64.0/18.)

# allow kube_service_addresses and kube_pods_subnet network
openstack port set 5662a4e0-e646-47f0-bf88-d80fbd2d99ef --allowed-address ip-address=10.233.0.0/18 --allowed-address ip-address=10.233.64.0/18
openstack port set e5ae2045-a1e1-4e99-9aac-4353889449a7 --allowed-address ip-address=10.233.0.0/18 --allowed-address ip-address=10.233.64.0/18

If all the VMs in the tenant correspond to Kubespray deployment, you can "sweep run" above with:

openstack port list --device-owner=compute:nova -c ID -f value | xargs -tI@ openstack port set @ --allowed-address ip-address=10.233.0.0/18 --allowed-address ip-address=10.233.64.0/18

Now you can finally run the playbook.

The external cloud provider

The in-tree cloud provider is deprecated and will be removed in a future version of Kubernetes. The target release for removing all remaining in-tree cloud providers is set to 1.21.

The new cloud provider is configured to have Octavia by default in Kubespray.

  • Enable the new external cloud provider in group_vars/all/all.yml:

    cloud_provider: external
    external_cloud_provider: openstack
    
  • Enable Cinder CSI in group_vars/all/openstack.yml:

    cinder_csi_enabled: true
    
  • Enable topology support (optional), if your openstack provider has custom Zone names you can override the default "nova" zone by setting the variable cinder_topology_zones

    cinder_topology: true
    
  • If you are using OpenStack loadbalancer(s) replace the openstack_lbaas_subnet_id with the new external_openstack_lbaas_subnet_id. Note The new cloud provider is using Octavia instead of Neutron LBaaS by default!

  • Enable 3 feature gates to allow migration of all volumes and storage classes (if you have any feature gates already set just add the 3 listed below):

    kube_feature_gates:
    - CSIMigration=true
    - CSIMigrationOpenStack=true
    - ExpandCSIVolumes=true
    
  • If you are in a case of a multi-nic OpenStack VMs (see kubernetes/cloud-provider-openstack#407 and #6083 for explanation), you should override the default OpenStack networking configuration:

    external_openstack_network_ipv6_disabled: false
    external_openstack_network_internal_networks:
    - ""
    external_openstack_network_public_networks:
    - ""
    
  • Run source path/to/your/openstack-rc to read your OpenStack credentials like OS_AUTH_URL, OS_USERNAME, OS_PASSWORD, etc. Those variables are used for accessing OpenStack from the external cloud provider.

  • Run the cluster.yml playbook