ktk is a command line tool to use many kubeconfig files simultaneously in different terminal tabs.

Originally written for kitty, ktk can also be used to manage tmux and wezterm tabs.

It is possible to customize the name and the color of the tabs for each cluster, to search quickly in thousands of namespaces, with a cache file.

When ktk open a new tab, you go directly to a working directory specific to the cluster and the namespace.

ktk can easily manage dozens of clusters with thousands of namespaces.

Example of the use of kitty tabs

Fuzzy search in ktk

Prerequisites

ktk needs kubectl and a terminal multiplexor like kitty, wezterm or tmux to work. And of course one or more Kubernetes clusters to manage.

Install

Copy the config.sample.yaml configuration file to the ~/.config/ktk.yaml file

Customize the file to declare your Kubernetes clusters.

Add the option allow_remote_control yes in the configuration file kitty.conf

And launch kitty with the -1 option to have only one kitty instance.

Configuration file ~/.config/ktk.yaml

Global section

In this section we declare the global settings of the application.

global:
  kubetmp: "/run/user/1000/.kubeconfig"
  separator: "::"
  completion:
    file: "/home/user/.kube/tkcompleted"
    maxage: 43200
  tabprefix: "☸>>"
  gradient:
    name: Spectral
    reverse: false
    darken: true
  oh-my-posh:
    file: "/home/user/.config/oh-my-posh/theme.json"

• kubetmp: folder where temporary kubeconfig files are copied.
• separator: separation between namespace and cluster name in the cache file and in the search.
• completion:
  • file: name of the cache file for the namespace search.
  • maxage: Duration of cache validity in seconds.
• tabprefix: prefix in the tab name, e.g. ☸>>kube-system::prod.
• gradient:
  • name: choose your gradient colors from Colorous.
  • reverse: reverse color gradient.
  • darken: if true, darkens the color of the inactive tab, otherwise inverts the color.- oh-my-posh: (optional).
  • file: oh-my-posh configuration file path.

Common settings for clusters

To avoid re-entering identical information for each cluster, it is possible to use the anchor system of the yaml file.

.workdir: &workdir
  path: "/home/user/kubernetes/deploy"
  subdir: "."
  prefixns: ""

.kubeconfig: &kubeconfig
  path: "/home/user/.kube/konfigs"
  file: "default"

• workdir: working folder when working in a namespace.
• path: main path.
• subdir: namespace specific path.
• prefix: prefix to remove from the namespace name. Example: prod- for a namespace like prod-my-app. So the path will be /home/user/kubernetes/deploy/my-app and not /home/user/kubernetes/deploy/prod-my-app.
• kubeconfig:
  • path: folder containing the kubeconfig file of the cluster.
  • file: name of the kubeconfig file.

Clusters section

Here are the specific settings for each cluster.

clusters:
  - name: prod
    workdir:
      <<: *workdir
      subdir: "prod_conf"
    kubeconfig:
      <<: *kubeconfig
      file: "prod"

Changing bashrc or zshrc

It is not possible to change the directory from an executable, so the program will display the export and cd commands which will be evaluated by eval.

if kubedir=$(~/.local/bin/ktk --evaldir); then
  eval "$(echo $kubedir)"
  source <(kubectl completion zsh)
  source <(stern --completion zsh)
  # ...
else
  export KUBECONFIG=/dev/null
fi
source <(ktk completion zsh)

Here I take the opportunity to load the automatic completion tools only if necessary.

[ ~/.config/ktk.yaml -nt ~/.config/oh-my-posh/theme.json ] && ktk -O

Since ktk v0.20.0, automatic update of oh-my-posh theme with ktk.

Usage

ktk 0.20.0
Kitty Tab for Kubeconfig

Usage: ktk [OPTIONS] [namespace::cluster]

Arguments:
  [namespace]  Namespace to operate on

Options:
  -c, --config <FILE>            Sets a custom config file [default: /home/gauthier/.config/ktk.yaml]
  -f, --force                    Force reconstruct cache of namespace
  -n, --noscan                   Do not reconstruct cache of namespace
  -C, --cluster                  Search only in current cluster like kubens (alias kubens="ktk -t -C")
  -l, --list-clusters-colors     List kube clusters with tabs colors in config file
  -L, --list-clusters-names      List kube clusters names in config file
  -O, --oh-my-posh-json          Update oh-my-posh json config file
  -s, --subfilter <subfilter>    Pre-filter on a subset of value with a regexp.
  -w, --wait                     disable timeout for namespaces search
  -t, --tab                      Change namespace without change tab (like kubens)
  -d, --debug                    Record debug event in log file
  -e, --evaldir                  Show in stdout workdir of current cluster
      --completion <completion>  Output shell completion code for the specified shell [possible values: bash, elvish, fish, powershell, zsh]
  -h, --help                     Print help (see more with '--help')
  -V, --version                  Print version

Examples:
  $ ktk kube-system::production
  $ ktk -t -C kube-system

To open a new tab with the prod cluster context for the default namespace :

# ktk default::prod

A new tab with the name ☸>>default::prod is opened with the correct kubeconfig context. If the command is run again, the focus will be on the tab that already has the same name.

The first time the command is run, ktk will scan all the clusters to get the list of namespaces and store this information in the cache.

To force the expiration of the cache (which by default lasts maxage seconds), especially when I have just created a new namespace, I just have to choose the -f option like this:

# ktk -f new-namespace::prod

If the name of the namespace is not complete, ktk opens a dialog to ask to choose in the list of possible namespaces, the one you want to reach.

  test5::prod
  temp-stage-ns::dev
> test1-namespace::sandbox
 3/32
> test

Special features

Due of the difference between kitty, wezterm and tmux, the behaviour of ktk differs between these 3 tools.

Tmux

Tmux does not support multiple tab coloring. Configuring the gradient color of the tabs in the config file has no effect.

Wezterm

Themes

Unlike kitty, it's not possible with wezterm to pass tab colors as parameters, but it is possible to read the ktk configuration file directly and create the color palette to be applied to tabs based on the Kubernetes context name.

local function get_cluster_ktk()
	local clusters = {}
	local file = io.open(wezterm.home_dir .. "/.config/ktk.yaml", "r")
	if file == nil then
		return
	end
	local content = file:read("*all")
	file:close()

	local data = wezterm.serde.yaml_decode(content)
	for _, v in pairs(data.clusters) do
		table.insert(clusters, v.name)
	end
	return clusters
end

local function get_gradient_ktk()
	local file = io.open(wezterm.home_dir .. "/.config/ktk.yaml", "r")
	if file == nil then
		return
	end
	local content = file:read("*all")
	file:close()

	local data = wezterm.serde.yaml_decode(content)
	return data.global.gradient.name
end

local clusters_kube = get_cluster_ktk()
local gradient_colors = wezterm.color.gradient({ preset = get_gradient_ktk() }, #clusters_kube)

The result is a color chart of the size of the number of clusters.

Workspaces

ktk manages wezterm workspaces as independent environments. It is possible to open a tab with an identical Kubernetes context (same namespace and same cluster) in 2 different workspaces.