kubectl config set-context accepted a namespace that does not exist

kubectl config set-context --namespace writes whatever string you give it - no API check. One typo and every later kubectl command silently targets a namespace that is not there, returning empty lists instead of errors. Catch it with kubectl config view --minify, fix it with kubens [real-name] (which validates), and prefer kubens for namespace switches going forward.

The error

Context "[ctx]" modified. (but the namespace "[typo]" does not exist - every kubectl call now targets it)

What to do

1. Spot the bad value:

kubectl config view --minify -o jsonpath='{.contexts[0].context.namespace}{"\n"}'

Expected: Prints the misspelled namespace.

1. Confirm it is not real:

kubectl get namespace [typo] 2>&1 | head -2

Expected: NotFound.

1. Fix with validation:

kubens [real-name]

Expected: Prints Active namespace is "[real-name]"..

1. Sanity check:

kubectl get pods

Expected: Shows the real pods instead of an empty list.

When this applies

• kubectl returning empty lists right after a set-context --namespace
• typos in --namespace values
• scripts that write namespace values without checking

When it does NOT apply

• kubens refusing a name (that IS validation working)
• namespaces that exist but are empty

Works with

kubectl; kubens for the validated fix

helm/k9s showing nothing after a context edit

Same dangling namespace in the shared kubeconfig. Fix it once and every tool recovers.

Why it happens

The kubeconfig namespace field is free text; set-context performs no read against the cluster. kubens exists partly to close that gap by validating first.

Edge cases

• Automation should run kubectl get namespace [name] before writing it into a context.
• A namespace deleted after being set produces the same silent-empty symptom - re-validate on failure.

Resolved from

gh:ahmetb/kubectx#83 - https://github.com/ahmetb/kubectx/issues/83