Bug 1503243 - [RFE] OpenStack Kuryr support
Summary: [RFE] OpenStack Kuryr support
Keywords:
Status: CLOSED CURRENTRELEASE
Alias: None
Product: OpenShift Container Platform
Classification: Red Hat
Component: RFE
Version: unspecified
Hardware: Unspecified
OS: Unspecified
high
high
Target Milestone: ---
: 3.10.0
Assignee: Michał Dulko
QA Contact: Jon Uriarte
URL:
Whiteboard:
Depends On:
Blocks: 1503708 1503975 1504017
TreeView+ depends on / blocked
 
Reported: 2017-10-17 15:54 UTC by Antoni Segura Puimedon
Modified: 2019-04-16 17:57 UTC (History)
17 users (show)

Fixed In Version:
Doc Type: If docs needed, set a value
Doc Text:
undefined
Clone Of:
Environment:
Last Closed: 2019-04-16 17:57:27 UTC
Target Upstream Version:
Embargoed:


Attachments (Terms of Use)


Links
System ID Private Priority Status Summary Last Updated
Github openshift openshift-ansible pull 5574 0 None closed Initial Kuryr support 2020-10-23 11:01:02 UTC
Github openshift openshift-docs pull 5845 0 None closed Add Kuryr SDN documentation. 2020-10-23 11:01:16 UTC

Description Antoni Segura Puimedon 2017-10-17 15:54:02 UTC
Description of problem:

In OpenShift of OpenStack environments we should have the option to use OpenStack Kuryr as the Software Defined Networking solution that openshift-ansible deploys so that OpenShift Pods and services can consume the native OpenStack networking provided.

This integration should consist on roles that deploy the Kuryr-Controller as an OpenShift deployment and the CNI component as a DaemonSet. All the relevant kuryr options should be made available as options the user can place in the inventory and with sane defaults.

Steps to Reproduce:
1. place kuryr Neutron options in /etc/ansible/hosts alongside the rest of the BYO configuration
2. ansible-playbook playbooks/byo/config.yml -vvv


Actual results:
Not implemented

Expected results:
OpenShift successfully deploys and the kube-system namespace has the Kuryr controller Deployment and the Kuryr CNI daemonset

Additional info:

Comment 1 Scott Dodson 2017-11-06 16:37:15 UTC
Changes have been merged into master but note that target release is 3.8.0.

Comment 2 Wei Sun 2018-01-23 02:33:38 UTC
Hi Ofer,
Please verify this bug,thanks!

Comment 5 Michał Dulko 2018-03-13 17:13:01 UTC
The feature is merged, though meanwhile we needed to make some more fixes as well. I'm not sure about QA status of that. As development team we've recently verified that this works.

Comment 6 Tomas Sedovic 2018-05-04 13:16:07 UTC
This has been implemented. However, the namespace is openshift-infra not kube-system.

How to test:

1. Configure the Kuryr SDN in the openshift-ansible inventory all.yml:

    openshift_use_kuryr: True
    openshift_use_openshift_sdn: False
    openshift_master_open_ports:
    - service: dns tcp
      port: 53/tcp
    - service: dns udp
      port: 53/udp
    openshift_node_open_ports:
    - service: dns tcp
      port: 53/tcp
    - service: dns udp
      port: 53/udp
    use_trunk_ports: True
    os_sdn_network_plugin_name: cni
    openshift_node_proxy_mode: userspace
    openshift_hosted_manage_registry: false
    # # Kuryr needs to know the subnet you will be taking Floating IPs for the
    # loadbalancer services from.
    kuryr_openstack_public_subnet_id: cdbd1855-0b23-4ddc-8186-e10a963156b8
    openshift_kuryr_precreate_subports: 5
    openshift_kuryr_device_owner: compute:kuryr
    enable_kuryr_controller_probes: False
    enable_kuryr_cni_probes: False

2. Run the openshift-ansible/playbooks/openstack/openshift-cluster/provision_install.yml playbook

Verification:

1. The playbook must finish without errors.
2. Running `oc get pod -n openshift-infra` from a master node should list one CNI pod for every OpenShift node and a single kuryr-controller pod

Comment 7 Jon Uriarte 2018-05-31 08:08:59 UTC
Verified in openshift-ansible-3.10.0-0.53.0.git.0.53fe016.el7.noarch and openshift-ansible-playbooks-3.10.0-0.53.0.git.0.53fe016.el7.noarch.

Verification steps:
1. Configure the Kuryr SDN in the openshift-ansible inventory all.yml:

    openshift_use_kuryr: True
    openshift_use_openshift_sdn: False
    openshift_master_open_ports:
    - service: dns tcp
      port: 53/tcp
    - service: dns udp
      port: 53/udp
    openshift_node_open_ports:
    - service: dns tcp
      port: 53/tcp
    - service: dns udp
      port: 53/udp
    use_trunk_ports: True
    os_sdn_network_plugin_name: cni
    openshift_node_proxy_mode: userspace
    openshift_hosted_manage_registry: false
    # # Kuryr needs to know the subnet you will be taking Floating IPs for the
    # loadbalancer services from.
    kuryr_openstack_public_net_id: dd5a700a-a0bf-4e18-b6db-a59f4063f7b4
    openshift_kuryr_precreate_subports: 5
    openshift_kuryr_device_owner: compute:kuryr
    enable_kuryr_controller_probes: False
    enable_kuryr_cni_probes: False

2. Run the openshift-ansible openstack playbook (deployed 1 master, 1 infra and 2 app nodes)
3. Verify kuryr-cni daemon set and kuryr controller deployment config:
  [openshift@master-0 ~]$ oc -n openshift-infra get ds
  NAME           DESIRED   CURRENT   READY     UP-TO-DATE   AVAILABLE   NODE SELECTOR   AGE
  kuryr-cni-ds   4         4         4         4            4           <none>          16h
  [openshift@master-0 ~]$ oc -n openshift-infra get cm
  NAME           DATA      AGE
  kuryr-config   2         16h


4. Verify kuryr controller and cni pods are ready and running:
  [openshift@master-0 ~]$ oc get pod -n openshift-infra
  NAME                                READY     STATUS    RESTARTS   AGE
  bootstrap-autoapprover-0            1/1       Running   0          16h
  kuryr-cni-ds-bcvrp                  1/1       Running   0          16h
  kuryr-cni-ds-hnqw2                  1/1       Running   0          16h
  kuryr-cni-ds-jvn5x                  1/1       Running   0          16h
  kuryr-cni-ds-kjxnv                  1/1       Running   0          16h
  kuryr-controller-65c98f7444-vv5l8   1/1       Running   0          16h

Comment 9 Michał Dulko 2019-04-16 17:57:27 UTC
It seems so, yes.


Note You need to log in before you can comment on or make changes to this bug.