Nodecellar Auto-scale Auto-heal Blueprint

This blueprint deploys a demo wine store application that is based on nodejs and mongodb using Cloudify.

Compatibility

Tested with:

Cloudify 4.3.1

Pre-installation steps

Please note the following requirement for manager configuration.

This example requires configuration of multiple management (agent) networks upon installation of your Cloudify Manager.

The required name of the agent network is external, and should map to a public IP address. For example:

[centos@ip-10-10-4-47 ~]$ sudo grep "networks" /etc/cloudify/config.yaml 
    networks: {default: 10.10.4.47, external: 54.67.45.103}

Upload the required plugins:

Check the blueprint for the latest version of the plugin.

Install the relevant example network blueprint for the IaaS that you wish to deploy on:

Installation

On your Cloudify Manager, navigate to Local Blueprints select Upload.

Right-click and copy URL. Paste the URL where it says Enter blueprint url. Provide a blueprint name, such as nodecellar in the field labeled blueprint name.

Select the blueprint for the relevant IaaS you wish to deploy on, for example openstack.yaml from Blueprint filename menu. Click Upload.

After the new blueprint has been created, click the Deploy button.

Navigate to Deployments, find your new deployment, select Install from the workflow's menu. At this stage, you may provide your own values for any of the default deployment inputs.

Deployment Outputs

Once the workflow execution is complete, we can view the application endpoint by running:

$ cfy deployments outputs nodecellar

You should see an output like this:

Retrieving outputs for deployment nodecellar...
 - "endpoint":
     Description: Web application endpoint
     Value: http://10.239.0.18:8080/

Use the URL from the endpoint output and visit that URL in a browser. Play with the wine store application.

Autoscaling

Execute a benchmarking command line application to simulate an auto-scaling scenario.

$ ab -n 1000000 -c 200 http://10.239.0.18:8080/ # insert your URL instead of this one.

This will increase the number of requests to the application. As a result the CPU used by the node process on the nodejs_host VMs will spike above the scale up threshold. This metric is monitored by the Diamond plugin, and the Riemann auto-scale policy calls the scale workflow trigger.

Killing this command should cause the CPU to drop below the scale down threshold, and the application will scale down.

Note: this assumes that the VM is an appropriate flavor for the benchmarking tool to sufficiently challenge the VM. This particular configuration is with a t2.micro AWS instance.

There are a number of ways to verify the scaling. If you have Cloudify Premium, you will see the executions in Cloudify's UI. You can also execute cfy executions list for a CLI view:

$ cfy executions list
Listing all executions...

Executions:
+--------------------------------------+-------------------------------+------------+---------------+--------------------------+-------+------------+----------------+------------+
|                  id                  |          workflow_id          |   status   | deployment_id |        created_at        | error | permission |  tenant_name   | created_by |
+--------------------------------------+-------------------------------+------------+---------------+--------------------------+-------+------------+----------------+------------+
| 10e5a704-6c97-43f0-bf84-cb3c3b5cf9e5 | create_deployment_environment | terminated |      nodecellar     | 2017-05-01 00:00:00.000  |       |  creator   | default_tenant |   admin    |
| a1dcbc8f-ae02-4fc7-80a6-1201a706b72b |            install            | terminated |      nodecellar     | 2017-05-01 00:00:00.000  |       |  creator   | default_tenant |   admin    |
| 53d9c07e-7340-4860-91c5-d402877be341 |             scale             | terminated |      nodecellar     | 2017-05-01 00:05:00.000  |       |  creator   | default_tenant |   admin    |
| 2dd15f7b-78ed-46db-9084-de8d0345ff3e |             scale             |  started   |      nodecellar     | 2017-05-01 00:10:00.000  |       |  creator   | default_tenant |   admin    |
+--------------------------------------+-------------------------------+------------+---------------+--------------------------+-------+------------+----------------+------------+

Simulate auto-healing

You can simulate a failed host by stopping or suspending a running nodejs_host VM. The Riemann failed host policy will recognize the lack of system cpu metrics reporting on the host and will trigger the heal workflow.

Uninstallation

Navigate to the deployment and select Uninstall. When the uninstall workflow is finished, select Delete deployment.

$ cfy uninstall --allow-custom-parameters -p ignore_failure=true nodecellar

bkobagana / nodecellar-auto-scale-auto-heal-blueprint Goto Github PK