Skip to content

Latest commit

 

History

History
122 lines (82 loc) · 4.56 KB

defining-a-custom-metric-87e657e.md

File metadata and controls

122 lines (82 loc) · 4.56 KB

Defining a Custom Metric

Define your own metrics to scale applications based on your requirements.

Context

As an alternative to the standard metrics provided by the Application Autoscaler, you can also define custom metrics such as memory consumed in megabytes or in percentage, response time, and throughput for scaling. Custom metrics offer more flexibility, so that you can scale applications based on your own requirements.

To use custom metrics, your application needs to perform the following tasks:

  1. Report a custom metric to the Application Autoscaler by defining a policy.

  2. Emit custom metric details using the HTTP API at regular intervals.

The following example shows a policy with a custom metric:

{
   "instance_min_count":1,
   "instance_max_count":4,
   "scaling_rules":[
      {
         "metric_type":"jobqueue",
         "breach_duration_secs":60,
         "threshold":100,
         "operator":">=",
         "cool_down_secs":120,
         "adjustment":"+1"
      }
   ]
}

Note:

The metric type used for custom metrics must not be any of the standard metric types.

Tip:

We recommend a minimum duration of one minute between successive emissions of a custom metric.

As part of the binding process, the Application Autoscaler service instance provides necessary credentials to emit custom metrics.

The generation of these credentials is based on the credential type parameter in the service binding. The supported credential types are binding-secret and x509. If no credential type is provided, binding-secret is used as the default credential type.

Procedure

  1. Bind your application with the Application Autoscaler service instance using a policy that contains custom metrics scaling rule.

    sample policy.json with credential Type as x509

    Sample Code:

    
{
    "instance_min_count":1,
    "instance_max_count":4,
    "scaling_rules":[
        {
        "metric_type":"jobqueue",
        "breach_duration_secs":60,
        "threshold":100,
        "operator":">=",
        "cool_down_secs":120,
        "adjustment":"+1"
        }
    ],
    "credential-type": "x509"
}

    bind the service instance with the scaling policy

    Sample Code:

    cf bind-service <application> <application-autoscaler-service-instance> -c policy.json

  2. Read custom metrics binding credentials from the application environment.

    Binding Credentials for mTLS

    Sample Code:

    
"custom_metrics": {
      "mtls_url": "https://autoscaler-metrics-mtls.cf.<landscape>.hana.ondemand.com",
     }


**Authenticate with X.509 Certificate:**

Your application requires base url, certificate and a private key for mutual TLS authentication \\\(mTLS\\\).

-   Use the base URL provided in the `mtls_url` of the service binding.

-   Use the X.509 certificate and private key from the paths provided in the environment variables `CF_INSTANCE_CERT` and `CF_INSTANCE_KEY` respectively.


> ### Note:  
> The environment variables `CF_INSTANCE_CERT` and `CF_INSTANCE_KEY` are not shown in the cockpit or using the `cf env` command. They are only visible inside the running application container.

**Authenticate with Basic Authentication \(Deprecated, do not use\):**

> ### Note:  
> The X.509 certificate and private key pair are valid for 24 hours. At least 20 minutes before expiration, they are regenerated and new files replace the existing files. Make sure that your code reloads the X.509 certificate and private key pair if they are expired. See [Using Instance Identity Credentials](https://docs.cloudfoundry.org/devguide/deploy-apps/instance-identity.html) in the Cloud Foundry Documentation.

`url`, `username`, and `password` in the application environment are provided for backwards compatibility with the HTTP basic authentication scheme and should not be used going forward. If your application still needs to authenticate with basic authentication, use credential-type binding-secret in the service binding and plan to migrate to mTLS \(\`x509\`\) authentication until 2025-01-23.
  1. Push custom metrics at an interval of your choice using APIs. For the REST API specification, see Application Autoscaler custom metrics API.