Guides on StormForge Documentation

Try a sample application on your own cluster

Mon, 01 Jan 0001 00:00:00 +0000

The Showcase application is a sample application we developed to create load, enabling you to experiment with Optimize Live on your own cluster: generate, view, and apply recommendations, observe how our machine learning rightsizes workloads, and set values to generate recommendations based on your organizational needs.

This application creates a single Kubernetes Deployment (workload) that generates variable load using one CPU core for 24 hours. Optimize Live collects metrics on this workload and generates recommendations to rightsize it.

Try the StormForge sandbox

Mon, 01 Jan 0001 00:00:00 +0000

If you don’t have a cluster of your own, you can check out Optimize Live running in a view-only sandbox.

The sandbox contains two clusters, each running the same simulated e-commerce application.

This guide walks you through a comparison of optimized and non-optimized workloads, showing you how Optimize Live machine learning rightsizes your workloads based on the metrics it collects.

Log in to the sandbox

To preview Optimize Live (with view-only privileges) without having to install anything, go to the sandbox landing page.

Manage workloads

Mon, 01 Jan 0001 00:00:00 +0000

Topics:

Adding workloads

Optimize Live automatically detects when the following types are added to a cluster:

Any Kubernetes Workload object (CronJob, DaemonSet, Deployment, ReplicaSet, StatefulSet) and HPA
Workloads scaled by a KEDA-owned HPA (ScaledObjects in keda.sh apiGroup)

When Optimize Live detects a new workload:

Within about 2 minutes, Optimize Live starts generating preliminary recommendations for the workload and updating the UI with workload and recommendation details. Preliminary recommendations cannot be applied automatically, but can be applied on demand. Optimize Live generates these preliminary recommendations for the duration of the workload’s learning period.
When a workload’s learning period is complete, Optimize Live generates recommendations that can be deployed either automatically or on demand.

Related information:

Monitoring the health of Optimize Live

Mon, 01 Jan 0001 00:00:00 +0000

Status metrics

StormForge Agent components expose health and status metrics that are valuable for monitoring the health of Optimize Live. Health and status metrics are different from the workload metrics Optimize Live needs to create request and limit recommendations. (See the Security FAQ for the list and descriptions of the workload metrics.)

Health and status metrics for each StormForge component are listed in the table below, under the component that produces them.

Configure applier alerts

Mon, 01 Jan 0001 00:00:00 +0000

The StormForge Applier can send an HTTP webhook notification each time it successfully applies a recommendation to a workload. You can configure one or more endpoints to receive these alerts, with optional custom payload formatting and bearer token authentication. Use this to connect StormForge to your existing observability or incident management workflows.

How alerts work

When the Applier applies a patch, it sends an HTTP POST to each configured endpoint. The request includes the header X-Event-Type: recommendation applied. The body contains either the full JSON recommendation object (default) or a custom payload that you can define using a Go template.

How to use Optimize Live and Argo CD together

Mon, 01 Jan 0001 00:00:00 +0000

Optimize Live and Argo CD are fully compatible — Argo CD just needs to be configured to understand that StormForge is operating in the environment.

After you install Optimize Live, add the following system-level reconcile configuration option to the argocd-cm ConfigMap:

data:
  resource.customizations.ignoreDifferences.all: |
    managedFieldsManagers:
    - stormforge

For more information on the underlying Kubernetes feature that enables managers like ArgoCD and StormForge to co-function, see:

Server-Side Apply details in the Kubernetes docs
Diffing Customization in the Argo CD docs

StormForge Kustomize export

Mon, 01 Jan 0001 00:00:00 +0000

For infrastructure workflows fully automated by a CI pipeline and Kustomize, an alternative option for deployment of recommended resource settings may be to incorporate StormForge’s recommendations into the CI pipeline.

Because Kustomizations (kustomization.yaml files) often serve as the entrypoint for application definitions in tools like Flux CD, tying into Kustomize is frequently the easiest way to automate resource settings if an organization wants or needs to do that during the CI/CD process, rather than managing these settings with a controller.

Import recommendation data into APM tools

Mon, 01 Jan 0001 00:00:00 +0000

By creating a Prometheus scrape job, Optimize Live exports recommendation data as Prometheus metrics, which you can import into other application performance management (APM) tools.

To define a scrape job to get recommendation data:

Edit the prometheus-server ConfigMap in the prometheus namespace:
```
kubectl edit configmap -n prometheus prometheus-server
```
Add the following scrape job to the scrape_configs: section of the ConfigMap.
- Replace SF_CLIENT_ID with the clientID value in your cluster credential file.

Transparent proxy configuration

Mon, 01 Jan 0001 00:00:00 +0000

If you use a transparent proxy, you need to configure StormForge to trust the proxy’s root certificate authority (CA). To do this, mount the proxy’s root CA certificate using a volume, then use additionalEnvVars to configure StormForge to use the mounted bundle.

Prerequisites

Helm version 3.14.0 or later (required for the --reset-then-reuse parameter)

Steps

Copy the YAML below into a new file named, for example, transparent-proxy-values.yaml. This example assumes the required root CA is available in a ConfigMap, but any core/v1 Volumes type can be used (see the related link below).

Manage secrets with a third-party credential manager

Mon, 01 Jan 0001 00:00:00 +0000

By default, Optimize Live uses Helm to create and update its credentials Secret. However, if you would like to use a third-party credential manager (for example, to rotate secrets), you can choose to disable Helm management of the credentials secret.

When Helm management of the credentials secret is disabled, you must:

Create the stormforge-system namespace before you install the agent.
Create the stormforge-agent-auth secret, and populate it with the required credentials, before you install the agent.
Maintain the stormforge-agent-auth secret yourself, manually or with a third-party credential manager.

Note

If you are not using a third-party credential management tool, you do not need to follow the steps in this guide. Instead, follow the steps in the Install topic. You still have to create an access credential (as described in the Install topic), but StormForge will create and manage the secret for you.

Optimize custom workloads using CRDs

Mon, 01 Jan 0001 00:00:00 +0000

When to use this guide

Use this guide if your organization defines workloads with CRDs, for example:

Custom workloads unique to your organization’s platform.
Third-party operators such as GitHub’s Actions Runner Controller (ARC) or Apache Spark applications.

By default, StormForge Optimize Live supports standard Kubernetes workload types such as Deployments, StatefulSets, and DaemonSets, as well as common third-party workload types, such as Argo Rollouts.

Optimize Live also supports custom workload types defined by Custom Resource Definitions (CRDs). If your CRD represents the workload itself, you can configure StormForge to recognize and optimize it using the workloadResourceTypes Helm parameter.

Optimize workloads owned by custom operators using CRDs

Mon, 01 Jan 0001 00:00:00 +0000

When to use this guide

Use this guide if your organization manages workloads with custom operators and you want greater control over owner identification and patch application.

In Kubernetes, operators—or custom controllers—can manage workloads. For example, a custom resource might own and manage a Deployment workload. StormForge automatically detects these ownership relationships.

When another resource (an operator) owns a workload, StormForge typically targets this owner for patching rather than the workload. This way, changes persist and aren’t overwritten by the operator’s reconciliation loop.

StormForge MCP server

Mon, 01 Jan 0001 00:00:00 +0000

The StormForge CLI includes an MCP server for connecting AI agents to StormForge.

This guide covers configuration and basic usage, using Claude as an example.

Prerequisites

Make sure the StormForge CLI is installed and you are logged in:

stormforge login

Configure your AI agent

This example uses Claude, but the configuration pattern applies to other MCP-compatible agents such as ChatGPT and Gemini.

Add the following to your agent’s MCP configuration. For Claude, that can be ~/.claude.json:

Subscribe through AWS Marketplace

Mon, 01 Jan 0001 00:00:00 +0000

With pay-as-you-go billing, you can subscribe to Optimize Live and get billed through your AWS account. Billing is based on an hourly rate, per the number of requested CPU cores for each cluster connected to your Optimize Live account. There is no upfront or minimum cost.

See the AWS Marketplace - Buyers Guide for details on the AWS Marketplace and the Optimize Live product page for Optimize Live pricing.

Subscribing to Optimize Live

Navigate to the Optimize Live product page in the AWS Marketplace.