GitHub - lukasdietrich/glance-k8s: Kubernetes Extension for https://github.com/glanceapp/glance (original) (raw)

An extension widget for the amazing Glance dashboard.

Features

This extension features two widgets to integrate Kubernetes resources into your dashboard.

Kubernetes Nodes

Setup

widgets:

• type: extension url: http://glance-k8s/extension/nodes allow-potentially-dangerous-html: true cache: 1s

Kubernetes Applications

Setup

widgets:

• type: extension url: http://glance-k8s/extension/apps allow-potentially-dangerous-html: true cache: 1s

  parameters:

  Parameters are sent to the extension as query parameters.

  Since there is no browser or proxy between glance and glance-k8s there should not be a size limit for the parameters,

  apart from default buffer sizes.

  But keep that in mind, when supplying large strings.

  Show only workloads, that match an expression.

  See: https://github.com/expr-lang/expr

  Environment:

  namespace Namespace of the workload

  name Name of the workload

  annotations Map of annotations

  show-if: | namespace != "kube-system" and ("glance/hide" not in annotations || annotations["glance/hide"] != "true")

  You can also supply multiple expressions, which will evaluate to a logical conjunction (AND).

  show-if:

  - |

  namespace != "kube-system"

  - |

  ("glance/hide" not in annotations || annotations["glance/hide"] != "true")

Customization / How it works

Kubernetes has a lot of moving parts, which makes it a little tricky to find all the installed applications.

The extension iterates over workloads (Deployment, StatefulSet and DaemonSet), services and ingresses in all namespaces.

Then it tries to match workloads to services and services to ingresses using their specified selectors to find an ingress for applications. Since configurations can become very complex, it might not be able to find the right ingress, if more than one exists. For most cases however, it should just work.

Finally the workloads are grouped into applications, which belong together. If you do not annotate workloads, every workload is assumed to be an application.

You can annotate workloads to group them into applications and customize their appearance on the dashboard. If the workload has an ingress, you may annotate the ingress as well.

apiVersion: apps/v1 kind: Deployment metadata: annotations: # Display Name (default: TitleCase(workload.Name)) glance/name: Glance

# Icon (default: di:kubernetes)
# The shorthands `di:` and `si:` are supported similar to glance.
glance/icon: di:glance
# glance/icon: https://example.org/glance.png

# Link to the application (default: Ingress with shortest path. First found in order: Main Workload > Dependencies)
glance/url: https://glance.example.org

# Open links on the same tab (default: false)
glance/same-tab: true

# Description
glance/description: My fancy dashboard

# Identifier for an application to group workloads.
# This should be annotated on the "main" workload of an application.
glance/id: glance

# Identifier of the main workload of the same app.
# This should be annotated on all workloads other than the "main" one of the same application.
# If multiple workloads have the same parent but there is no workload annotated with `glance/id`
# the first found will be promoted to be the "main" workload.
glance/parent: glance

Configuration

The extension is configured via environment variables. All variables are optional.

About the response cache

A single dashboard pageload causes Glance to fire one HTTP request per widget to glance-k8s. With many per-category widgets (each filtering on a distinct glance/id), the requests fan out in parallel and each one would otherwise trigger fresh cluster-wide List() calls for deployments, statefulsets, daemonsets, services, ingresses, and HTTPRoutes — which on larger clusters exhausts client-go's default 5 QPS limit and produces multi-second cold-loads.

To avoid this, cluster-wide List() responses are cached in-process for a short, fixed TTL. Concurrent callers for the same resource are collapsed onto a single in-flight fetch via singleflight, and errors are not cached so a transient apiserver failure does not lock the cache for the full TTL.

Installation

Glance itself provides a container image, but no official helm chart yet. Until then, this repository contains charts for both upstream glance as well as the glance-k8s extension.

Updates of glance are tracked using a self-hosted Renovate Bot running on github actions.

Use the latest versions instead

export GLANCE_VERSION=v0.8.3 export GLANCE_K8S_VERSION=v0.1.3

See https://helm.sh/docs/helm/helm_install/

You can provide a values file to the install command using -f values.yaml

helm install glance oci://ghcr.io/lukasdietrich/glance-k8s/chart/glance:${GLANCE_VERSION} helm install glance-k8s oci://ghcr.io/lukasdietrich/glance-k8s/chart/glance-k8s:${GLANCE_K8S_VERSION}

Values

The default values can be found in their respective chart folder:

• glance
• glance-k8s