Skip to content

Organize the output

Kubesplit names each generated file after the resource it contains and groups files by namespace. This guide shows how to control that layout. For the exhaustive rules, see File naming and layout.

Default layout: prefixed and grouped by namespace

By default every file is named:

<order-prefix>--<kind>--<name>.yml

Cluster-wide resources (Namespace, ClusterRole, ClusterRoleBinding, …) go to the root of the output directory; namespaced resources go into a subdirectory named after their namespace.

kubesplit --input test-assets/source/all-in-one.yml \
          --output out \
          --no-quotes-preserved \
          --clean-output-dir

out
├── apps-demo
│   └── 05--rolebinding--example-ns-demo-developer-binding.yml
├── apps-integration
│   └── 05--rolebinding--example-ns-integration-developer-binding.yml
├── ingress-controllers
│   ├── 03--serviceaccount--traefik-ingress-controller.yml
│   ├── 11--configmap--traefik-conf.yml
│   ├── 12--persistentvolumeclaim--traefik-acme.yml
│   ├── 20--deployment--traefik-ingress-controller.yml
│   ├── 30--service--traefik-ingress-endpoint.yml
│   ├── 30--service--traefik-web-ui.yml
│   └── 31--ingress--traefik-web-ui.yml
├── 00--namespace--apps-demo.yml
├── 00--namespace--apps-integration.yml
├── 00--namespace--ingress-controllers.yml
├── 01--clusterrole--example-node-viewer.yml
├── 01--clusterrole--example-traefik-ingress-controller.yml
├── 02--clusterrolebinding--example-node-viewer-developer.yml
└── 02--clusterrolebinding--example-traefik-ingress-controller.yml

3 directories, 16 files

Why the number prefixes?

The 00, 01, 20, 30… prefixes are apply-order weights derived from the resource kind. Namespaces come first (00), RBAC next (01–05), config and storage in the middle (10–13), workloads (20–25), then networking (30–42). Applying the files in lexical order (which is what kubectl apply -f out/ --recursive does) therefore respects the usual dependency order.

The full prefix table lives in File naming and layout.

Drop the prefixes with --no-resource-prefix

If you don't need an apply order (for example when files feed a GitOps tool that resolves ordering itself), use -p/--no-resource-prefix:

kubesplit --input test-assets/source/all-in-one.yml \
          --output out \
          --no-quotes-preserved \
          --no-resource-prefix \
          --clean-output-dir

out
├── apps-demo
│   └── rolebinding--example-ns-demo-developer-binding.yml
├── ingress-controllers
│   ├── configmap--traefik-conf.yml
│   ├── deployment--traefik-ingress-controller.yml
│   ├── ingress--traefik-web-ui.yml
│   ├── persistentvolumeclaim--traefik-acme.yml
│   ├── serviceaccount--traefik-ingress-controller.yml
│   ├── service--traefik-ingress-endpoint.yml
│   └── service--traefik-web-ui.yml
├── clusterrole--example-node-viewer.yml
├── clusterrolebinding--example-node-viewer-developer.yml
└── namespace--apps-demo.yml

...

The namespace grouping is unaffected — only the NN-- prefix is removed.

Next steps