Overview

This page provides guidance for administrators on how to manage Knative on an existing Kubernetes cluster. Knative administrators install and configure both or either of the Serving and Eventing components along with default or preferred plugins.

Administrators can use Knative to provide developers with a simple experience for interacting with clusters and deploying applications. In this model, developers primarily interact with Knative resources like Services, Brokers, and Triggers. Because Knative can interoperate with core Kubernetes objects, developers can also use existing Kubernetes tools such as pods, services, networking, identity, and storage where needed. Developers looking to further simplify the deployment experience can define functions with the Knative Functions programming model. The following illustration shows the roles of administrators and developers in this model:

---
config:
  theme: redux
  layout: dagre
  look: classic
---
flowchart LR
 subgraph Knative["**Knative**  "]
    direction LR
        Serving["Serving"]
        Eventing["Eventing"]
  end
 subgraph Plugins["**Plugins**  "]
    direction LR
        net-istio["Istio"]
        net-contour["Contour"]
        net-gateway-api["Gateway API"]
        event-kafka["Kafka"]
        event-rabbitmq["RabbitMQ"]
        event-nats["NATS"]
  end
    Dev(["**Developers**"]) --> dev-acts["Develops and manages"]
    dev-acts --> Serving & Eventing
    Admin(["**Administrators**"]) --> admin-acts["Installs and configures"]
    admin-acts --> Knative & Plugins
    Serving --> net-impl["Controls"]
    net-impl --> net-istio & net-contour & net-gateway-api
    Eventing --> event-impl["Controls"]
    event-impl --> event-kafka & event-rabbitmq & event-nats
    dev-acts@{ shape: text}
    admin-acts@{ shape: text}
    net-impl@{ shape: text}
    event-impl@{ shape: text}
    style Serving fill:#D5D5D5,color:#000000
    style Eventing fill:#F0DBDB,color:#000000
    style net-istio fill:#D5D5D5,color:#000000
    style net-contour fill:#D5D5D5,color:#000000
    style net-gateway-api fill:#D5D5D5,color:#000000
    style event-kafka fill:#F0DBDB
    style event-rabbitmq fill:#F0DBDB
    style event-nats fill:#F0DBDB
    style Dev fill:#EFB769,color:#000000
    style dev-acts fill:transparent
    style Admin fill:#94C6C1,color:#000000
    style admin-acts fill:transparent
    style Knative text-align: left
    style Plugins text-align: left
    style net-impl fill:transparent
    style event-impl fill:transparent

As a cluster administrator, your responsibilities include managing the Kubernetes environment, installing cluster-wide components, and enabling developers to deploy applications on the cluster. Knative aims to simplify developer tasks, while aligning with existing management tools and processes.

Knative includes a plugin system to integrate with existing infrastructure in the cluster, enabling Knative resources such as Routes and Brokers to be implemented using one of multiple underlying suppliers. For example, a Knative Eventing app can deliver events to a Broker that triggers a function based on the received event. In a testing cluster, the delivery might use an in-memory option, while a staging or production environment might use a cloud-provided Kafka service.

Of particular interest to cluster administrators is that Knative supports customizable default values on the parameters defined in resource YAML files. These configurations reduce the amount of environment configuration tasks developers need to consider.

Installation decisions

See the Installation roadmap for prerequisites and installation steps. Your first installation decision is whether to use a YAML-based installation or use the Knative Operator. The Knative Operator is a custom controller that extends the Kubernetes API to install Knative components. If you just need to get acquainted with Knative at this time, you can install the quickstart.

The method you use to install Knative is not permanent and you can install clusters differently depending on the situation. Although transitioning between installation methods on one cluster is possible, new installations on separate clusters is the better-tested and officially supported approach.

Upgrades

Administrators are generally responsible for performing upgrades to cluster infrastructure, apps, and services. Knative is designed and tested for continuous operation during upgrades and rollbacks, allowing you to:

• Upgrade or revert the Knative components while it is serving traffic, rather than needing a maintenance window.
• Downgrade by one Knative version. Downgrades work provided that no applications have used new features since the last upgrade.

Securing Knative

Knative resources are namespaced. Knative adheres to the Kubernetes model of namespace-based isolation that lets you manage development teams and resources by assigning them to namespaces. You may also grant developers access to additional resources related to their namespace in other services, such as observability, logs, metrics, tracing, and dashboards.

Namespaces can also isolate boundaries for tooling such as logs, metrics, tracing, CI/CD integrations, and dashboards. The extent of this isolation depends on both the enforcement strategy and how consistently teams adhere to namespace boundaries.

You can optimize and enforce isolation involving namespaces using standard Kubernetes mechanisms, including:

• Role-Based Access Control (RBAC)
• Resource Quotas
• Network Policies
• Pod Security Standards

Configurations

Knative configurations are performed by the following methods:

• Editing YAML manifests and applying with the kubectl tool

  Modify resource definitions directly, including labels, annotations, and field values. You can use Kubernetes features such as OPA and Kyverno to enforce specific values on a resource type, or use ConfigMaps in plugin installations to set values at the cluster level.

• Using ConfigMaps

  Store and manage configuration data as key-value pairs. ConfigMaps are frequently used to tune platform-wide behavior. Most of the Knative ConfigMaps are in the knative-serving and knative-eventing namespaces. Their settings apply to all the relevant Knative components in all namespaces.

• Using the Knative Operator

  Some platform-wide settings can be managed declaratively using the Knative Operator, installed with the kn Knative CLI plugin. You can manage the operator without using the kn CLI. The kn CLI manages only operator installations.

  For more information, see Installing CLI tools

Knative uses Kubernetes YAML manifests to define and configure system components. These manifests include core resources, custom resource definitions (CRDs), and extensibility features. As with Kubernetes, these configuration resources are declarative and can be managed using the kubectl CLI tool or with continuous delivery tools.

The following sections provide an overview of the current configuration resources of interest to Administrators. You can edit these configurations using kubectl; Knative installs empty ConfigMaps with these names onto the cluster.

Serving configurations

Configuration	ConfigMap	Description
Default configurations	config-defaults	Default resource values such as performance, hardware, and storage settings.
Deployment resources	config-deployment	Kubernetes deployment resources that back Knative services.
Domain names	config-domain	Configure and publish domains.
High-availability	NA	Configure ensure that APIs stay operational if a disruption occurs.
Garbage collection	config-gc	Disable and enable collection and set retention time values.
Ingress gateway	config-istio	For new clusters, you can configure your own gateway and underlying service.
Istio authorization	NA	Grant authorization to your deployed Knative services.
Namespace exclusion from webhook	NA	For performance concerns during an upgrade.
Rollout duration for revisions	config-network	Adjust rollout durations to accommodate longer request queues.
Security - Certificates	config-certmanager	Describes how to manage automatic certificate provisioning.
Security - Encryptions	config-network	Provides links to procedures for encrypting external domains, the local cluster, and system internal.

Eventing configurations

Configuration	ConfigMap	Description
Broker defaults	config-br-defaults	Specify your own broker class and channel, or use the default MTChannelBasedBroker Broker class and the ConfigMap of channel defaults.
Broker features (Kafka)	config-kafka-features	Configure options for Broker interactions with Apache Kafka clusters.
Channel defaults	default-ch-webhook	Default configurations and labels to use for the channel.
Channel defaults (Kafka)	kafka-channel	Defines how KafkaChannel instances are created. Requires that KafkaChannel custom resource definitions (CRD) are installed.
Event source defaults	config-ping-defaults	Configure the PingSource default resources and the maximum data size for CloudEvents it produces.
KEDA Autoscaling of Kafka Resources	config-kafka-features	Configure how KEDA scales a KafkaSource, trigger, or subscription. Note: This feature is is Alpha pre-release.
Sugar Controller	config-sugar	Configure the Sugar controller, which reacts to label configurations to produce or control eventing resources. See also Knative Eventing Sugar Controller.