Skip to content
Kedify Docs

For the complete documentation index and AI-optimized content, see /llms.txt. All pages support markdown format via .md extension or Accept: text/markdown header.

Kedify Agent

For the complete documentation index and AI-optimized content, see /llms.txt. All pages support markdown format via .md extension or Accept: text/markdown header.

The Kedify Agent is responsible for establishing a secure connection between your Kubernetes cluster and the Kedify platform. It handles the installation and lifecycle management of KEDA and the HTTP Scaler on your clusters, providing robust scaling capabilities and telemetry insights. The Kedify Agent can manage a large number of installations across multiple clusters using secure gRPC connections to communicate with the Kedify Dashboard, ensuring the seamless collection and display of telemetry data.

Key functionalities of the Kedify Agent include:

  • Installation Management: The agent manages the installation and versioning of KEDA and the HTTP Scaler using the Helm package manager. It ensures that the correct versions of these components are deployed and maintained across your clusters.

  • Telemetry Collection: The agent includes a telemetry collector that scrapes metrics from KEDA, the HTTP Scaler, and opted-in Kubernetes workloads. This telemetry data provides users with valuable insights into their configurations, scaling behaviors, and container resource recommendations, which are displayed in the Kedify Dashboard under Insights section.

  • Secure Communication: The Kedify Agent establishes a secure, outbound gRPC connection to the Kedify Service, ensuring that all data transmitted between your clusters and the Kedify platform is secure.

  • Custom Resource Management: The agent manages Kedify custom resources, providing autowiring and comprehensive management of the HTTP Scaler. For more information on custom resources, refer to the Kedify Custom Resource Definitions.

The Kedify Agent can be installed through helm chart or using a single, unique kubectl apply command generated and provided by Kedify via the dashboard. This command sets up the agent with all necessary configurations to connect to the Kedify platform and manage your KEDA and HTTP Scaler installations efficiently.

The Kedify Agent relies on the Helm package manager to manage and install versioned KEDA images.

Configuration

Section titled “Configuration”

Full set of configuration options is available in Kedify Agent chart values. Here are some notable options:

  • agent.features.podResourceProfilesEnabled: Enables the controller for PodResourceProfile to be started by Kedify Agent. The feature is further described on its own docs. By default, it is true. Env var equivalent: PRP_ENABLED.

  • agent.features.prpRequiresAnnotatedPods: If enabled, the controller for PodResourceProfile requires the target pods (where container resources should be modified) to have annotation prp.kedify.io/reconcile=enabled on them. This is a performance optimization so that more pod events can be ignored by the controller. Default: true. Env var equivalent: PRP_REQUIRES_ANNOTATED_PODS.

  • agent.features.podResourceAutoscalersEnabled: Enables the controller for Pod Resource Autoscaler. Default is false; set it to true to enable PRA. Env var equivalent: PRA_ENABLED.

  • agent.features.distributedScaledObjectsEnabled: Enables the controller for DistributedScaledObject used by Multi-Cluster Scaling. Starting with Kedify Agent v0.5.0, default is false and you need to set it to true to activate the feature. Env var equivalent: DSO_ENABLED.

  • agent.features.distributedScaledJobsEnabled: Enables the controller for DistributedScaledJob used by Multi-Cluster Scaling. Starting with Kedify Agent v0.5.0, default is false and you need to set it to true to activate the feature. Env var equivalent: DSJ_ENABLED.

  • agent.features.scalingPoliciesEnabled: Enables the controller for Scaling Policy. Default is true; set it to false to disable the controller. Env var equivalent: SCALING_POLICY_ENABLED.

  • agent.features.recommendationsForLabeledNamespaces: Limits Kubernetes container utilization scraping to opted-in namespaces. Default: true. Env var equivalent: METRICS_FOR_LABELED_NAMESPACES_ENABLED.

  • agent.metricsWatchNamespace: Comma-separated list of namespaces where Kubernetes container utilization metrics should be collected. When agent.features.recommendationsForLabeledNamespaces=true, explicit namespaces in this list are added to the labeled namespaces. When label-gated collection is disabled, an unset or empty value means all namespaces. Env var equivalent: METRICS_WATCH_NAMESPACE.

  • agent.kedifyProxy.clusterWide: Enables the cluster-wide deployment of kedify-proxy, shared across all namespaces in the cluster. This variable should be set to true to configure a single instance of kedify-proxy in the KEDA installation namespace for centralized scaling. The cluster-wide setup is supported only for configurations using Istio’s VirtualService. For more details, see Kedify Proxy.

  • agent.kedifyProxy.globalValues: Set of options to configure kedify-proxy deployments. See Kedify Proxy helm chart for more details. This can also be fine-tuned per namespace through agent.kedifyProxy.namespacedValues.${NAMESPACE}.

Container Metrics for Resource Recommendations

Section titled “Container Metrics for Resource Recommendations”

Kedify Agent can collect container CPU and memory utilization from the Kubernetes Metrics API and send it to Kedify for resource recommendations. To keep collection explicit, namespace label filtering is enabled by default.

Opt in a namespace by adding the insights.kedify.io/scrape label with any value except false, disabled, or paused:

Terminal window
kubectl label namespace production insights.kedify.io/scrape=enabled

Pause or disable collection for a namespace by changing the label value:

Terminal window
kubectl label namespace production insights.kedify.io/scrape=paused --overwrite

You can also remove the label to opt the namespace out:

Terminal window
kubectl label namespace production insights.kedify.io/scrape-

If you manage the namespace list centrally using GitOps, set agent.metricsWatchNamespace on the Kedify Agent Helm chart. Explicit namespaces from this value are still collected when label-gated collection is enabled. To return to the older cluster-wide behavior, set agent.features.recommendationsForLabeledNamespaces=false; with that setting, agent.metricsWatchNamespace controls the scrape scope and an unset or empty namespace list means all namespaces.

Container utilization metrics require the Kubernetes Metrics API, usually provided by metrics-server or a compatible adapter. The Kedify Agent Helm chart includes the required RBAC for namespace discovery and workload metric collection.

API Key

Section titled “API Key”

The Kedify Agent uses an API key to authenticate with the Kedify platform. This key is generated by the Kedify platform and provided to the agent during installation. There could be multiple valid API keys defined per Organization.

The API key can be rotated in the Kedify Dashboard in the Organization -> API Keys section. When rotated, it is the responsibility of the user to update the API key in the Kedify Agent configuration. The key is stored in a Kubernetes Secret as a base64 encoded string.

First, run the new API key through the base64 function:

Terminal window
key="kfy_971fac53610186bc26891e08b73076c297552073f57c4c6fa6c563c6821d4384"
echo -n "$key" | base64 -w 0 ; echo ""
a2Z5Xzk3MWZhYzUzNjEwMTg2YmMyNjg5MWUwOGI3MzA3NmMyOTc1NTIwNzNmNTdjNGM2ZmE2YzU2M2M2ODIxZDQzODQ=

Then, update the API key in the Secret. Either in your gitops repository or directly in the cluster:

Terminal window
kubectl edit secret kedify-agent -n keda
apiVersion: v1
kind: Secret
type: Opaque
metadata:
  name: kedify-agent
  namespace: keda
data:
  apikey: a2Z5Xzk3MWZhYzUzNjEwMTg2YmMyNjg5MWUwOGI3MzA3NmMyOTc1NTIwNzNmNTdjNGM2ZmE2YzU2M2M2ODIxZDQzODQ=