With App Autoscaler, you can scale an app in your TAS for VMs deployment based on a custom metric. This tutorial walks you through the steps.
In a VMware Tanzu Application Service for VMs (TAS for VMs) deployment, Autoscaler can automatically scale apps based on custom metrics. The following table describes the main components involved in this workflow and how they correspond to steps in this tutorial.
Component | Description | Related Tutorial Steps |
---|---|---|
App | The app must emit custom metrics that are created with the open-source tool, Prometheus. This tutorial includes a sample Spring app that does that. | Review the Sample App and Push the Sample App |
Metric Registrar | The Metric Registrar is a component of TAS for VMs that allows app developers to export custom app metrics to the logging system. It has its own CLI plugin. | Register a Custom Metrics Endpoint |
App Autoscaler | App Autoscaler is a service integrated with Apps Manager that automatically scales apps in your environment based on app metrics or a schedule. See the Scaling an App Using App Autoscaler topic for more information. | Create and Autoscaling Rule and Trigger Scaling |
This tutorial requires the following:
The sample app code is in the pivotal-cf/metric-registrar-examples GitHub repository. It is a Spring app with a simple UI that includes several buttons to call different endpoints. Some of these endpoints are instrumented to produce metrics.
You can see what the UI looks like in the Push the Sample App section, which includes a screenshot. The following sections provide some details about the code.
You can view the app dependencies in the build.gradle file:
dependencies {
implementation('io.micrometer:micrometer-registry-prometheus')
implementation('org.springframework.boot:spring-boot-starter-actuator')
implementation('org.springframework.boot:spring-boot-starter-security')
implementation('org.springframework.boot:spring-boot-starter-web')
testImplementation('org.springframework.boot:spring-boot-starter-test')
testImplementation('org.springframework.security:spring-security-test')
}
The dependencies include the Micrometer Prometheus library, which does the following:
/actuator/prometheus
in a format supported by the Metric Registrar.The Spring Security dependency exposes the endpoints so they can be reached by Metric Registrar.
This section describes how the app is instrumented. Instrumentation refers to how metrics have been added for a particular function.
See the ExampleController.java file from the sample app code:
The ExampleController
class includes a MeterRegistry
object, which is passed and set in the constructor.
private MeterRegistry registry;
private AtomicLong custom;
public ExampleController(MeterRegistry registry) {
this.registry = registry;
this.custom = new AtomicLong(0L);
}
It also includes a custom
variable that is initialized in the constructor. In the customMetric handler, this variable gets passed to the registry and incremented or decremented.
public ResponseEntity<String> customMetric(@RequestParam(value="inc", defaultValue="") String increment) {
AtomicLong customGauge = registry.gauge("custom", this.custom);
if (!"".equals(increment)) {
customGauge.incrementAndGet();
} else {
customGauge.decrementAndGet();
}
Note: The App Autoscaler only scales on a gauge, or metric, that can go up and down. The standard metrics of CPU, disk, HTTP throughput, and HTTP latency are all gauges.
To push the sample app:
Open a terminal window and clone the git repository that contains the sample app:
git clone git@github.com:pivotal-cf/metric-registrar-examples.git
Navigate to the app directory:
cd metric-registrar-examples/java-spring-security
Build the app:
./gradlew build
Push the app with a random route:
cf push --random-route
After the push
command finishes, locate the app URL in the routes:
section of the output. See the following example output:
Waiting for app to start... Uploaded droplet (60.5M) Uploading complete Cell 333e7fdf-806e-424d-b3a0-78967ecb6d28 stopping instance 6f345835-8beb-48a5-b578-921f5de442c6 name: tutorial-example requested state: started routes: tutorial-example-random-route.cfapps.io last uploaded: Wed 28 Aug 11:02:33 PDT 2019
In a browser, navigate to the app URL. The app UI looks like what you see in the following image:
It has the following buttons:
When you want to your app to emit custom metrics, you register the app as a metric source with the Metric Registrar.
To register a custom metrics endpoint for the app:
Install the Metric Registrar CLI:
cf install-plugin -r CF-Community "metric-registrar"
Register the metrics endpoint of the app. Because the app dependencies include the Micrometer Prometheus library, there is automatically a metrics endpoint at /actuator/prometheus
. Run the following command and replace APP-NAME
with the name of your app.
cf register-metrics-endpoint APP-NAME /actuator/prometheus
Install the Log Cache CLI. Log Cache is a feature in TAS for VMs that lets you filter and query app logs. For more information, see Example Uses of the Log Cache CLI Plugin.
cf install-plugin -r CF-Community "log-cache"
View the app metrics using the following command. The --follow
flag appends output as metrics are emitted.
cf tail APP-NAME --envelope-class metrics --follow
The output looks similar to the following example. The custom
metric displays in the output.
Retrieving logs for app tutorial-example in org sandbox / space development as example@user... 2019-08-28T09:17:56.28-0700 [tutorial-example/1] GAUGE cpu:0.289158 percentage disk:135716864.000000 bytes disk_quota:1073741824.000000 bytes memory:399979315.000000 bytes memory_quota:2147483648.000000 bytes 2019-08-28T09:17:56.50-0700 [tutorial-example/0] GAUGE custom:1.000000
Note: If you do not see output similar to the previous exampple, Metric Registrar might not be enabled in your Ops Manager installation. Contact your platform operator to confirm.
The App Autoscaler component is integrated with Apps Manager. This is where you can create rules.
To create an autoscaling rule for the app and custom metric:
Navigate to the app in Apps Manager.
Click Enable Autoscaling.
Click Manage Autoscaling.
Modify the Instance Limits.
App Autoscaler keeps instance counts within a range defined by minimum and maximum values, or instance limits.
1
.5
.Click Apply Changes.
Create an autoscaling rule.
App Autoscaler increases or decreases instance counts based on how a current metric compares with configured Scale up and Scale down thresholds.
Custom
.2
. This and the following value are examples for the purposes of demonstrating the feature in this tutorial.5
.custom
. This is the name of the metric specified in the app code.Click Save.
Now that you have pushed an app that emits a custom metric and configured autoscaling rules, you can trigger a scaling action. App Autoscaler scales the app when the custom metric goes above or below the threshold specified in the scaling rule.
To trigger scaling:
Navigate to the web UI of the app. Use the same URL from Push the Sample App.
Click Increment Custom gauge enough times to bring the custom metric over the threshold of 5
that you set in the scaling rule. You can check the value of the custom
metric using the See Metrics button.
Monitor the app page in Apps Manager for about two minutes. App Autoscaler will begin to scale the app. It adds one instance at a time until it reaches the Maximum instance limit of 5
.
Now that you have completed this tutorial, try emitting custom metrics and creating scaling rules with your own app. Review the resources listed in the Overview section to learn more. Once you have instrumented your app to emit custom metrics, you can follow the steps outlined in this tutorial to scale based on those metrics.