Skip to content

Smart filters

Every resource list has a search box. By default it's a plain text search — every word you type has to appear somewhere in the row (name, namespace, cluster, status, node, images or labels). Start your query with a / and it becomes a smart filter: structured clauses for status categories, numeric comparisons, labels and ages that narrow thousands of pods down to the ones that matter in one line:

/status:crash ns:prod cpu>500m restarts>3

The moment you type /, Kubus autocompletes the keys that make sense for the kind you're looking at, including live values (your namespaces, clusters and nodes).

How queries work

  • The query starts with /; everything after it is the smart filter.
  • Clauses are separated by spaces and ANDed together.
  • A clause is either free text (nginx) or key:value / key>value / key<value.
  • OR alternatives with a comma: /status:crash,oom, /ns:dev,staging.
  • Negate with ! — /!ns:kube-system or /status:!running both work.
  • Quotes protect spaces: /name:"billing worker".
  • Everything is case-insensitive, and an unknown key just falls back to free text.

Universal keys

Key Matches Example
name: name contains /name:api
ns: / namespace: namespace contains /ns:prod
cluster: / ctx: cluster contains /cluster:staging
label: exact label, or key presence; * globs /label:app=nginx, /label:team
annotation: exact annotation, or key presence /annotation:owner=platform
age> / age< resource age (s, m, h, d, w) /age>7d, /age<30m
status: status text or a category alias /status:degraded

status: understands aliases beyond the literal status text:

  • crash → CrashLoopBackOff, oom → OOMKilled
  • error → errors, failures and backoffs of any flavour
  • unhealthy / healthy — anything not fully up (works for pods, workloads and nodes)
  • degraded / progressing — workloads with fewer ready replicas than desired
  • completed — Succeeded pods and complete Jobs

Kind-specific keys

Key Kinds Example
restarts> Pods /restarts>5
node: Pods /node:worker-1
image: Pods /image:redis
cpu> / mem> Pods, Nodes /cpu>500m, /mem>1Gi, /cpu>80%
ready: Pods, Nodes, workloads /ready:false
replicas> Deployments, StatefulSets, … /replicas>3
type: Services /type:lb, /type:np
reason: / message: Events /reason:BackOff

Absolute quantities, not just percentages

cpu> and mem> take real Kubernetes quantities (250m, 1.5, 512Mi, 2Gi). Percentages (cpu>80%) compare against capacity where Kubus knows it — node utilisation, for example.

Filtering by label

Next to the search box, the Labels dropdown filters server-side by label selector. It lists every label key and key=value pair present in the rows — tick as many as you like (they're ANDed together), or type a raw selector like env!=prod and press Enter.

Every row also shows its labels as chips in the Labels column — hover to see them all, and click a chip to add it to the label filter.

Saved views remember your filter

A smart filter is part of the page URL, so saved views capture it — save /status:unhealthy in prod once and it's one click from the nav drawer forever.

See also