🐶 k9s Quick Reference

k9s [flags]
k9s -n my-namespace

k9s is a terminal dashboard for Kubernetes. It shows your cluster live (pods, deployments,
nodes, logs) and lets you describe, edit, shell into, restart or delete things with single
keystrokes, instead of typing long kubectl commands. It uses the same kubeconfig as kubectl,
so it can do whatever your account is allowed to do. Be careful: one wrong key on the
wrong context can delete something real.

brew install k9s                          # macOS and Linux (Homebrew)
sudo snap install k9s                     # Linux (snap)
go install github.com/derailed/k9s@latest # from source
winget install k9s   |   choco install k9s   |   scoop install k9s     # Windows
Or download a release binary from github.com/derailed/k9s/releases and put it on your PATH.
docker run --rm -it -v ~/.kube/config:/root/.kube/config quay.io/derailed/k9s
k9s version                               # check what you have

k9s                                       # current context and namespace
k9s -n kube-system                        # start in a namespace
k9s -A                                    # all namespaces (--all-namespaces)
k9s --context prod-cluster                # a specific context
k9s --kubeconfig ~/.kube/other            # a specific kubeconfig
k9s -c deployments                        # start on a view (--command; pods, svc, nodes ...)
k9s --readonly                            # disable every write action (safe for browsing)
k9s --headless                            # hide the header
k9s --logoless                            # hide the logo
k9s --crumbsless                          # hide the breadcrumbs
k9s -l debug                              # log level (logs go to a file, see k9s info)
k9s info                                  # show where config, logs and skins are stored
k9s --help

Header (top)      Context, cluster, user, k9s and Kubernetes versions, CPU and memory, plus key hints
Main table        The resources in the current view. The selected row is highlighted
Breadcrumbs       Where you are (for example: pods > containers)
Command line      Appears at the bottom when you press : or /

j / k or Down / Up     Move down / up             g / G    Jump to top / bottom
Ctrl-F / Ctrl-B        Page down / up             Enter    Drill into the selected item
Esc                    Go back / close a prompt   ?        Show every key for the current view
Ctrl-A                 List all available resource aliases (like :pod, :svc)
:q  or Ctrl-C          Quit

:pods  (:po)                  :deployments  (:deploy)         :services  (:svc)
:namespaces  (:ns)            :nodes  (:no)                   :configmaps  (:cm)
:secrets  (:sec)              :ingresses  (:ing)              :statefulsets  (:sts)
:daemonsets  (:ds)            :replicasets  (:rs)             :jobs   :cronjobs  (:cj)
:persistentvolumes  (:pv)     :persistentvolumeclaims  (:pvc) :events  (:ev)
:serviceaccounts  (:sa)       :roles  :rolebindings           :clusterroles  :clusterrolebindings
:context  (:ctx)              Switch cluster context
:pulses                       Cluster health summary charts
:xray deploy                  Tree of a resource and everything it owns (deploy, pods, svc ...)
:popeye                       Cluster sanitizer report (needs Popeye support)
:alias                        List all aliases
:pods /nginx                  A view with a filter already applied
:pods kube-system             A view in a specific namespace
Use the Tab key to autocomplete a command. Custom resources (CRDs) appear as views too.

:ns  then Enter on a row     Use that namespace
0                            Show all namespaces
1 ... 9                      Jump to a favorite namespace (numbers shown in the header)
:ctx  then Enter on a row    Switch to another cluster context
u  (in the namespaces view)  Use the selected namespace
Double-check the context name in the header before you do anything destructive.

/error               Filter the table by text (regular expression; press Enter to lock it)
/!error              Invert: hide rows that match
/-l app=web          Filter by label selector
/-f web              Fuzzy filter
Esc                  Clear the filter
/  in logs and YAML  Search the text; n and N jump to the next and previous match

Enter    Drill down (deployment > pods > containers)
d        Describe (like kubectl describe)
y        View the YAML
e        Edit the resource in your $EDITOR (saves back to the cluster)
l        View logs (pods and containers)
p        Logs of the previous container instance (after a crash)
s        Shell into the container (pods). On deployments/statefulsets: scale
a        Attach to the container
Shift-F  Port-forward (opens a dialog for the ports)
f        Show the active port-forwards (in the port-forward list)
r        Restart the workload (deployments, statefulsets, daemonsets)
Ctrl-D   Delete (asks for confirmation)
Ctrl-K   Kill: delete right now, no grace period (asks for confirmation)
Space    Mark or unmark a row (act on several at once)
Ctrl-Space  Mark a range
Ctrl-\   Clear all marks
Shift-J  Jump to the owner of a resource (pod to its deployment)
o        Show which node a pod runs on
u        Show what uses a secret or configmap (some views)
x        Decode a secret (in the secrets view)
t        Trigger a cronjob (in the cronjobs view)
Ctrl-Z   Toggle showing only resources with errors (some versions)
Which keys exist depends on the resource type and k9s version: press ? to see the exact list.

Shift-N   Sort by name                Shift-A   Sort by age
Shift-C   Sort by CPU                 Shift-M   Sort by memory
Shift-S   Sort by status              Shift-R   Sort by ready
Press the same key again to flip ascending and descending.
Ctrl-W    Toggle wide columns         Ctrl-E    Toggle the header

l        Open logs for the selected pod or container
0        Show logs since the beginning (tail)
1        Last 1 minute      2  Last 5 minutes      3  Last 15 minutes
4        Last 30 minutes    5  Last 1 hour
s        Toggle autoscroll  w   Toggle line wrap    t  Toggle timestamps
f        Full screen        c   Copy to the clipboard
m        Mark a position (for reading later)        Ctrl-S   Save the log to a file
/        Search text in the log
p        Switch to the previous container's logs
Esc      Back to the pod list
For pods with several containers, choose the container first, or use the container view.

Select the pod, then press s. k9s picks the container (or asks). Exit with exit or Ctrl-D.
Some minimal images have no shell. In that case use a debug container:
kubectl debug -it POD --image=busybox --target=CONTAINER

:pulses            Charts of pods, deployments, jobs and events over time
:xray deploy       See an entire deployment tree with statuses
:popeye            Scan the cluster for misconfigurations and wasteful settings
:nodes then d      Describe a node: capacity, taints, conditions, pods
:events            Recent cluster events (sort with Shift-A to see the newest)
:pods then Shift-C or Shift-M    Find the CPU and memory hogs
Metrics (CPU and memory columns) need metrics-server installed in the cluster.

Usually ~/.config/k9s/   (or $XDG_CONFIG_HOME/k9s, or ~/Library/Application Support/k9s on macOS)
config.yaml          Global settings: refresh rate, readOnly, UI options, log settings
aliases.yaml         Your own command shortcuts
hotkeys.yaml         Your own key bindings to jump to views
plugins.yaml         Your own actions that run external commands
views.yaml           Choose and order table columns per resource
skins/               Color themes
clusters/            Per-cluster settings (in some versions)

aliases:
  pp: v1/pods
  dep: apps/v1/deployments
  dc: apps/v1/deployments
Then type :pp to open the pods view.

hotKeys:
  shift-0:
    shortCut: Shift-0
    description: Show the pods view
    command: pods

plugins:
  stern:
    shortCut: Shift-L
    description: Tail logs with stern
    scopes:
      - pods
    command: stern
    background: false
    args:
      - --tail
      - 50
      - $NAME
      - -n
      - $NAMESPACE
Plugin variables include $NAMESPACE, $NAME, $POD, $CONTAINER and $CONTEXT. Restart k9s after editing.
Only add plugins you trust. They run commands on your machine with your cluster access.

k9s:
  refreshRate: 2                  # seconds between screen refreshes
  readOnly: false                 # true disables all edit, delete and scale actions
  noExitOnCtrlC: true             # avoid quitting by accident with Ctrl-C
  logger:
    tail: 200
    buffer: 5000
    sinceSeconds: 300
    textWrap: false
    showTime: false
Skins: copy a theme into ~/.config/k9s/skins/ and set ui.skin in config.yaml to its file name.

# Why is my pod crashing?
:pods          then / to filter by name
d              describe: read the Events at the bottom
l              view logs   (and p for the previous container's logs)

# Roll a deployment
:deploy        select it, press r to restart, or s to scale, or e to edit

# Follow the logs of everything a deployment runs
:deploy        Enter (shows its pods), select a pod, press l

# Find noisy or heavy pods
:pods          Shift-C (CPU) or Shift-M (memory)

# Peek inside a secret
:secrets       select it, press x to decode

# Reach an internal service from your laptop
:svc           select it, Shift-F, choose ports, then open http://localhost:PORT

# Check the whole cluster quickly
:pulses        then :nodes, :events

# Work on a different cluster without leaving k9s
:ctx           Enter on the context you want

# Look at one namespace only
:ns            Enter on it, or start with k9s -n my-namespace

# Safe browsing for a demo or on production
k9s --readonly

k9s vs kubectl and friends
kubectl        The exact, scriptable command line. Use it in scripts and CI
k9s            Live, interactive exploration and quick fixes
kubectx/kubens Quick context and namespace switching
stern          Tail logs from many pods at once
lens / headlamp   Graphical Kubernetes dashboards
kubectl get -w Simple live watch of a single resource type

"Unable to connect to the cluster"   Your kubeconfig or VPN is wrong: try kubectl get nodes first
Empty screen or no resources         Wrong namespace: press 0 for all namespaces, or use :ns
"Forbidden" messages                 Your account lacks permission (RBAC) for that resource
CPU and memory columns are blank     metrics-server is not installed or not ready
Edit does nothing or opens the wrong editor   Set $EDITOR or $KUBE_EDITOR (for example export EDITOR=vim)
Screen is garbled                    Use a terminal with 256 colors and a UTF-8 locale; try TERM=xterm-256color
Cannot shell into a pod              The image has no shell (scratch, distroless). Use kubectl debug instead
Keys behave differently              k9s versions differ. Press ? to see the keys for your version and view
Config changes ignored               Restart k9s. Check the config path with k9s info

    Press ? in any view. It always shows the exact keys that work for that resource and your version.

    Learn six things first: : (views), / (filter), d (describe), l (logs), s (shell), Esc (back).

    Check the context name in the header before you press Ctrl-D or Ctrl-K. Production and staging can look identical.

    Use k9s --readonly when you only want to look, or in shared sessions and demos.

    Press 0 to see all namespaces, and use / to filter, instead of scrolling.

    Sort by CPU or memory (Shift-C, Shift-M) to find heavy workloads fast, and by age (Shift-A) to find what changed recently.

    Add aliases, hot keys and plugins for the things you do every day, such as a stern log tailer.

    Use :xray and :pulses for the big picture, and :events to see what the cluster is complaining about.

    k9s is a client. It never changes anything you could not do with kubectl and your own permissions.

    Deleting a pod managed by a Deployment just recreates it. To stop it, scale the Deployment to 0 instead.