Using the Application Autoscaler to Manage Your Workload

Let’s prepare a scaling policy for the application. The scaling policy defines the minimum and maximum number of instances. It also defines the scaling rules based on the metric type, threshold, operator, and adjustment.

1. Clone the repository from GitHub (if you haven’t yet).

   1. Enter:

      git clone https://github.com/SAP-samples/cloud-foundry-runtime-learning-journey.git

   2. Navigate to the sample application directory:

      – cd cloud-foundry-runtime-learning-journey

2. Build the application.

   1. Enter:

      npm install

   2. Enter:

      cds build --production

3. Create a file cpu_scaling_policy.json with the following content:

   JSON

   Copy codeSwitch to dark mode

   12345678910111213141516171819202122
   {
     "instance_min_count": 1,
     "instance_max_count": 2,
     "scaling_rules": [
       {
       "metric_type": "cpuutil",
       "breach_duration_secs": 120,
       "threshold": 80,
       "operator": ">",
       "cool_down_secs": 300,
       "adjustment": "+1"
       },
       {
       "metric_type": "cpuutil",
       "breach_duration_secs": 120,
       "threshold": 80,
       "operator": "<",
       "cool_down_secs": 300,
       "adjustment": "-1"
       }
     ]
   }

   1. Enter the code from above.

      Let's understand the above scaling policy:

      • Scale out rule: If the CPU utilization is greater than 80% for two minutes (120 seconds), then add one instance to the application, provided the maximum instance count is not reached.
      • Scale in rule: If the CPU utilization is less than 80% for two minutes (120 seconds), then remove one instance from the application, provided the minimum instance count is not reached.

1. Log in to the Cloud Foundry environment using the Cloud Foundry CLI.

   1. Enter:

      cf login --sso

      Target the cf org and space, where <org> and <space> are the org and space names respectively:

      cf target -o <org> -s <space>

2. Push the application to Cloud Foundry without starting it.

   1. Enter:

      cf push -f manifest.yml --no-start

      Info: You used the CF manifest file (manifest.yml) to deploy the application on BTP Cloud Foundry. Another approach is to use a multitarget application (MTA). Check out the documentation for more details.

1. Create a service instance of Application Autoscaler with service instance name my-autoscaler-service, service plan standard, and service name autoscaler.

   1. Enter:

      cf create-service autoscaler standard my-autoscaler-service

1. Attach the autoscaler and scaling policy with the application.

   1. Enter:

      cf bind-service hello-cloud-foundry-srv my-autoscaler-service -c cpu_scaling_policy.json

2. Start the application.

   1. Enter:

      cf start hello-cloud-foundry-srv

3. Check if the application is in running state.

   cf app hello-cloud-foundry-srv

   Output:

   Code Snippet

   Copy codeSwitch to dark mode

   1234567891011121314151617
   # Output
   Showing health and status for app hello-cloud-foundry-srv in org ccc5375etrial / space name: hello-cloud-foundry-srv
   requested state: started
   isolation segment: trial
   routes: hello-cloud-foundry-srv.cfapps.us10-001.hana.ondemand.com
   last uploaded: Mon 02 Sep 12:20:35 CEST 2024
   stack: cflinuxfs4
   buildpacks:
   isolation segment: trial
   name version detect output buildpack name
   nodejs_buildpack 1.8.25 nodejs nodejs
   type: web
   sidecars:
   instances: 1/1
   memory usage: 256M
     state   since           cpu       memory     disk     details
   #0 running 2024-09-03T11:26:37Z 1.1% 70M of 256M 184.8M of 1G

   As shown in the output, the application is running with one instance at zero index.

4. Access our deployed application.

   1. Grab the application route from the output of the previous command. To do so, enter:

      Code Snippet

      Copy codeSwitch to dark mode

      1
      cf app hello-cloud-foundry-srv | grep routes

      In our case, the application route is: https://hello-cloud-foundry-srv.cfapps.us10-001.hana.ondemand.com. If you access the URL, you'll see the welcome message of the application, which confirms that the application is running

Now, your application is attached to the autoscaler service and scaling policy.

Let’s apply some CPU load on the application to see the scaling in action.

Luckily, your sample application offers an endpoint ’/useCPU’ to artificially increase high cpu usage on the application.

1. Open a browser, and issue an HTTP GET request to apply a CPU load on the application. Also, you are free to use any tool of your choice, for example, Postman for a HTTP client, or curl to send a HTTP request to the application.

   1. Enter:

      https://hello-cloud-foundry-srv.cfapps.us10-001.hana.ondemand.com/odata/v4/catalog/useCPU(action= 'start',duration='4')

      The output:

      JSON

      Copy codeSwitch to dark mode

      12345678
      {
        "@odata.context": "$metadata#Edm.String",
        "value": {
        "status": "started",
        "metric": "cpuutil",
        "service": "Application Autoscaler"
        }
      }

In this task, you'll verify if the application is scaled based on our scaling policy (defined in Task 1).

1. Check the application scaling status.

   1. Use the following command:

      cf app hello-cloud-foundry-srv

      Output:

      Code Snippet

      Copy codeSwitch to dark mode

      12345678910111213141516171819
      # Output
      Showing health and status for app hello-cloud-foundry-srv in org ccc5375etrial / space 
      name: hello-cloud-foundry-srv
      requested state: started
      isolation segment: trial
      routes: hello-cloud-foundry-srv.cfapps.us10-001.hana.ondemand.com
      last uploaded: Tue 03 Sep 17:30:56 CEST 2024
      stack: cflinuxfs4
      buildpacks:
      isolation segment: trial
      name             version   detect output   buildpack name
      nodejs_buildpack 1.8.25    nodejs          nodejs
      type: web
      sidecars:
      instances: 2/2
      memory usage: 256M
        state   since               cpu   memory       disk     details
      #0 running 2024-09-03T15:38:43Z 0.9% 69.4M of 256M 184.8M of 1G
      #1 running 2024-09-03T15:36:56Z 0.9% 71.1M of 256M 184.8M of 1G

      As shown in the output, the application is scaled out to two instances based on the scaling policy (defined in Task 1) by the Application Autoscaler service.

      Now, relax and enjoy the performance of your application using the Application Autoscaler service.

1. Unbind the autoscaler service from the application.

   1. Enter:

      cf unbind-service hello-cloud-foundry-srv my-autoscaler-service

2. Delete the autoscaler service instance.

   1. Enter:

      cf delete-service my-autoscaler-service -f

3. Delete the application.

   1. Enter:

      cf delete hello-cloud-foundry-srv -f