Adds cluster autoscale for ocp4 and cleans some descriptions (#370)

Erik M Jacobs

2019-04-19 d99a611241499470102a09121fece81bf9a636aa

Adds cluster autoscale for ocp4 and cleans some descriptions (#370)

* Adds cluster autoscale for ocp4 and cleans some descriptions

* switches to using k8s_facts for fetching machinesets

* Fixes node count in readme

 ansible/roles/ocp4-workload-cluster-autoscale/defaults/main.yml

New file
@@ -0,0 +1,6 @@
---
become_override: False
ocp_username: opentlc-mgr
silent: False

_project_request_message: "To provision Projects you must request access in https://labs.opentlc.com or https://rhpds.redhat.com"

 ansible/roles/ocp4-workload-cluster-autoscale/readme.adoc

New file
@@ -0,0 +1,118 @@
= ocp4-workload-cluster-autoscale - Sets up cluster autoscaling and creates machineautoscalers

== Role overview

* This role configures the cluster autoscaler and creates machine autoscalers
  for the default worker pool. It assumes a default cluster size of 3 workers
  and will allow the cluster to scale up to 9 workers. It consists of the
  following playbooks:
** Playbook: link:./tasks/pre_workload.yml[pre_workload.yml] - Sets up an
 environment for the workload deployment.
*** Debug task will print out: `pre_workload Tasks completed successfully.`

** Playbook: link:./tasks/workload.yml[workload.yml] - Used to set up autoscalers
*** Debug task will print out: `workload Tasks completed successfully.`

** Playbook: link:./tasks/post_workload.yml[post_workload.yml] - Used to
 configure the workload after deployment
*** This role doesn't do anything here
*** Debug task will print out: `post_workload Tasks completed successfully.`

** Playbook: link:./tasks/remove_workload.yml[remove_workload.yml] - Used to
 delete the workload
*** This role removes the infrastructure nodes (DANGER!!!)
*** Debug task will print out: `remove_workload Tasks completed successfully.`

== Review the defaults variable file

* This file link:./defaults/main.yml[./defaults/main.yml] contains all the variables you need to define to control the deployment of your workload.
* The variable *ocp_username* is mandatory to assign the workload to the correct OpenShift user.
* A variable *silent=True* can be passed to suppress debug messages.
* You can modify any of these default values by adding `-e "variable_name=variable_value"` to the command line

=== Deploy a Workload with the `ocp-workload` playbook [Mostly for testing]

----
TARGET_HOST="bastion.na311.openshift.opentlc.com"
OCP_USERNAME="shacharb-redhat.com"
WORKLOAD="ocp-workload-enable-service-broker"
GUID=1001

# a TARGET_HOST is specified in the command line, without using an inventory file
ansible-playbook -i ${TARGET_HOST}, ./configs/ocp-workloads/ocp-workload.yml \
    -e"ansible_ssh_private_key_file=~/.ssh/keytoyourhost.pem" \
    -e"ansible_user=ec2-user" \
    -e"ocp_username=${OCP_USERNAME}" \
    -e"ocp_workload=${WORKLOAD}" \
    -e"silent=False" \
    -e"guid=${GUID}" \
    -e"ACTION=create"
----

=== To Delete an environment

----
TARGET_HOST="bastion.na311.openshift.opentlc.com"
OCP_USERNAME="opentlc-mgr"
WORKLOAD="ocp4-workload-infra-nodes"
GUID=1002

# a TARGET_HOST is specified in the command line, without using an inventory file
ansible-playbook -i ${TARGET_HOST}, ./configs/ocp-workloads/ocp-workload.yml \
    -e"ansible_ssh_private_key_file=~/.ssh/keytoyourhost.pem" \
    -e"ansible_user=ec2-user" \
    -e"ocp_username=${OCP_USERNAME}" \
    -e"ocp_workload=${WORKLOAD}" \
    -e"guid=${GUID}" \
    -e"ACTION=remove"
----


== Other related information:

=== Deploy Workload on OpenShift Cluster from an existing playbook:

[source,yaml]
----
- name: Deploy a workload role on a master host
  hosts: all
  become: true
  gather_facts: False
  tags:
    - step007
  roles:
    - { role: "{{ocp_workload}}", when: 'ocp_workload is defined' }
----
NOTE: You might want to change `hosts: all` to fit your requirements


=== Set up your Ansible inventory file

* You can create an Ansible inventory file to define your connection method to your host (Master/Bastion with `oc` command)
* You can also use the command line to define the hosts directly if your `ssh` configuration is set to connect to the host correctly
* You can also use the command line to use localhost or if your cluster is already authenticated and configured in your `oc` configuration

.Example inventory file
[source, ini]
----
[gptehosts:vars]
ansible_ssh_private_key_file=~/.ssh/keytoyourhost.pem
ansible_user=ec2-user

[gptehosts:children]
openshift

[openshift]
bastion.cluster1.openshift.opentlc.com
bastion.cluster2.openshift.opentlc.com
bastion.cluster3.openshift.opentlc.com
bastion.cluster4.openshift.opentlc.com

[dev]
bastion.cluster1.openshift.opentlc.com
bastion.cluster2.openshift.opentlc.com

[prod]
bastion.cluster3.openshift.opentlc.com
bastion.cluster4.openshift.opentlc.com
----

 ansible/roles/ocp4-workload-cluster-autoscale/tasks/main.yml

New file
@@ -0,0 +1,23 @@
---

# Do not modify this file

- name: Running Pre Workload Tasks
  import_tasks: ./pre_workload.yml
  become: "{{ become_override | bool }}"
  when: ACTION == "create" or ACTION == "provision"

- name: Running Workload Tasks
  import_tasks: ./workload.yml
  become: "{{ become_override | bool }}"
  when: ACTION == "create" or ACTION == "provision"

- name: Running Post Workload Tasks
  import_tasks: ./post_workload.yml
  become: "{{ become_override | bool }}"
  when: ACTION == "create" or ACTION == "provision"

- name: Running Workload removal Tasks
  import_tasks: ./remove_workload.yml
  become: "{{ become_override | bool }}"
  when: ACTION == "destroy" or ACTION == "remove"

 ansible/roles/ocp4-workload-cluster-autoscale/tasks/post_workload.yml

New file
@@ -0,0 +1,9 @@
---
# Implement your Post Workload deployment tasks here


# Leave this as the last task in the playbook.
- name: post_workload tasks complete
  debug:
    msg: "Post-Workload Tasks completed successfully."
  when: not silent|bool

 ansible/roles/ocp4-workload-cluster-autoscale/tasks/pre_workload.yml

New file
@@ -0,0 +1,9 @@
---
# Implement your Pre Workload deployment tasks here


# Leave this as the last task in the playbook.
- name: pre_workload tasks complete
  debug:
    msg: "Pre-Workload tasks completed successfully."
  when: not silent|bool

 ansible/roles/ocp4-workload-cluster-autoscale/tasks/remove_workload.yml

New file
@@ -0,0 +1,31 @@
# vim: set ft=ansible
---
# Implement your Workload removal tasks here

- name: delete cluster autoscaler
  k8s:
    state: absent
    definition:
      apiVersion: "autoscaling.openshift.io/v1alpha1"
      kind: "ClusterAutoscaler"
      metadata:
        name: "default"

- name: get machine auto scalers
  k8s_facts:
    api_version: autoscaling.openshift.io/v1alpha1
    kind: MachineAutoscaler
    namespace: openshift-machine-api
  register: machine_auto_scalers

- name: delete machine autoscalers
  k8s:
    state: absent
    definition: "{{ item }}"
  with_items: "{{ machine_auto_scalers.resources }}"

# Leave this as the last task in the playbook.
- name: remove_workload tasks complete
  debug:
    msg: "Remove Workload tasks completed successfully."
  when: not silent|bool

 ansible/roles/ocp4-workload-cluster-autoscale/tasks/workload.yml

New file
@@ -0,0 +1,50 @@
# vim: set ft=ansible
---
# Implement your Workload deployment tasks here
- name: get current machinesets
  k8s_facts:
    api_version: machine.openshift.io/v1beta1
    kind: MachineSet
    namespace: openshift-machine-api
  register: machinesets_list

- name: create machine autoscaler for each machineset
  k8s:
    state: present
    definition:
      apiVersion: "autoscaling.openshift.io/v1alpha1"
      kind: MachineAutoscaler
      metadata: 
       name: "autoscale-{{ item.metadata.name }}"
       namespace: "openshift-machine-api"
      spec: 
       minReplicas: 1
       maxReplicas: 4
       scaleTargetRef: 
        apiVersion: "machine.openshift.io/v1beta1"
        kind: MachineSet
        name: "{{ item.metadata.name }}"
  with_items: "{{ machinesets_list.resources }}"

- name: create the cluster autoscaler
  k8s:
    state: present
    definition:
      apiVersion: "autoscaling.openshift.io/v1alpha1"
      kind: "ClusterAutoscaler"
      metadata:
        name: "default"
      spec:
        resourceLimits:
          maxNodesTotal: 9
        scaleDown:
          enabled: true
          delayAfterAdd: 10s
          delayAfterDelete: 10s
          delayAfterFailure: 10s

# Leave this as the last task in the playbook.
- name: workload tasks complete
  debug:
    msg: "Workload Tasks completed successfully."
  when: not silent|bool