This page describes the architecture of the node-manager module for Static nodes.
Module architecture
The following simplifications are made in the diagram:
- The diagram shows containers in different pods interacting directly with each other. In reality, they communicate via the corresponding Kubernetes Services (internal load balancers). Service names are omitted if they are obvious from the diagram context. Otherwise, the Service name is shown above the arrow.
- Pods may run multiple replicas. However, each pod is shown as a single replica in the diagram.
The Level 2 C4 architecture of the node-manager module and its interactions with other Deckhouse Platform (DP) components are shown in the following diagram:
Module components
Bashible is a key component of the Cluster & Infrastructure subsystem that enables the operation of the node-manager module. However, it is not part of the module itself, as it runs at the OS level as a system service. For Bashible details, refer to the corresponding documentation section.
The module managing Static nodes consists of the following components:
-
Bashible-api-server: A Kubernetes Extension API Server deployed on master nodes. It generates bashible scripts from templates stored in custom resources. When kube-apiserver receives a request for resources containing bashible bundles, it forwards the request to bashible-api-server and returns the generated result. For more details about bashible and bashible-api-server, refer to the corresponding documentation section.
-
Node-controller (Deployment): A controller that manages NodeGroup custom resources lifecycle. Node-controller performs the following operations:
- Manages NodeGroup custom resources lifecycle.
- Implements NodeGroup custom resources validating webhooks using the Validating Admission Controllers mechanism.
- Implements NodeGroup and Instance custom resources conversion webhooks.
- Cleans up Node resource labels and taints that remain after the bashible first run to initialize the node.
- Ensures draining a node.
- Applies labels, taints and annotations from the
spec.nodeTemplatesection of a NodeGroup custom resource to all Node resources belonging to it. - Calculates and updates a NodeGroup custom resource
statussubresource based on aggregated information obtained from the corresponding Node resources and infrastructure custom resources. - Sets the
spec.providerId = "static://"attribute for Static type Node resources if it is missing. - Manages the node update lifecycle: approving updates, handling node disruptions, draining a node and cleanup after a successful update.
The component includes:
- node-controller: Main container.
- kube-rbac-proxy: Sidecar container providing an RBAC-based authorization proxy for secure access to controller metrics.
-
Node-group-exporter (Deployment): A component that exports NodeGroup resource metrics in Prometheus format, containing information about the number of nodes in each node group: the total number, the number of nodes in the
Readystatus, the number of nodes in error, the minimum and maximum number of nodes in the group, etc.The component includes:
- node-group-exporter: Main container.
- kube-rbac-proxy Sidecar container providing an RBAC-based authorization proxy for secure access to exporter metrics.
-
Capi-controller-manager (Deployment): Core controllers from the Kubernetes Cluster API project. Cluster API extends Kubernetes to manage clusters as custom resources within another Kubernetes cluster.
The component includes:
- control-plane-manager: Main container.
- kube-rbac-proxy: Sidecar container providing an RBAC-based authorization proxy for secure access to controller metrics.
-
Caps-controller-manager (Deployment): CAPI Provider Static (CAPS), an implementation of a provider for declarative management of static nodes (bare-metal servers or virtual machines) in the Kubernetes Cluster API project. It operates as an extension to capi-controller-manager.
CAPS provides an additional abstraction layer over the existing DP mechanism for automatic configuration and cleanup of static nodes using scripts generated for each node group. The component is not tied to a specific cloud provider. For more details, refer to the
node-managerdocumentation. -
Fencing-agent (DaemonSet) and fencing-controller: Components that implement the fencing mechanism. The operation principles of both components are described in detail in the
spec.fencing.modeparameter description of the NodeGroup resource. For details on how the fencing mechanism handles different node types, refer to FAQ in thenode-managermodule documentation.
Module interactions
The module interacts with the following components:
-
Kube-apiserver:
- Works with Cluster API custom resources.
- Manages Node and NodeGroup resources.
- Authorizes metric requests.
-
Node filesystem:
/dev/watchdog: Sends signals to reset the Watchdog timer.
The following external components interact with the module:
-
Kube-apiserver:
- Executes validating and conversion webhooks of node-controller.
- Executes mutating and validating webhooks of capi-controller-manager.
- Forwards requests for bashible resources to bashible-api-server.
-
Prometheus-main:
- Collects metrics from
node-managermodule components.
- Collects metrics from
Architecture features specific to Static nodes
-
Users create and configure nodes in the following ways:
- Manually, using bashible scripts preconfigured in DP.
- Manually, with the following node handover to CAPS for automated management.
- Automatically, using CAPS.
-
Capi-controller-manager manages the lifecycle of the cluster and its nodes through higher-level custom resources, without directly provisioning infrastructure. It generates infrastructure-specific custom resources, leaving provisioning to the infrastructure provider, which is deployed by the specific cloud provider module (CAPS for static nodes).
- Caps-controller-manager: Component managing static nodes (partially, without provisioning).
- Static nodes can be used not only on bare metal, but in cloud as well. In that case, such a node is not managed by cloud-controller-manager, even if one of the cloud providers is enabled. Csi-driver is not installed on such nodes.
- Automatic node scaling is not supported.