Query resources in current namespace by default
ahmetb opened this issue · comments
This came up in #9, #14 and #16.
Right now (v0.3.x) we are querying ALL objects in ALL api groups (without any namespace limitation). This has several problems:
- fails to query api X, because user has access to X in namespace1, but not in namespace2
- takes too long to enumerate all APIs across all namespaces (#16)
From controller examples like Kubernetes Deployment, Knative, Agones, etc. it's fair to assume that resources don't often create child resources in other namespaces.
Proposed solution
When kubectl tree is querying child objects of a specified object:
- we should query only "namespaced" objects in the same namespace as the specified object as the new default
- we should offer
--all-namespacesoption to retain the current behavior (e.g. query all objects in all API groups –both namespaced and non-namespaced)
This addresses:
- slowness when used with large clusters with too many objects
- allow an escape hatch for the current behavior to still exist
- allow plugin to work when user has (full-) access to only some namespaces
It would likely work by refactorying query method to be namespace-scoped (query only namespaced APIs in the specified namespace) OR global-scoped (query both namespaced & non-namespaced APIs without specifying a namespace limitation, i.e. what we do today).
Anyone willing to take this on?
Anyone willing to take this on?
☝️ I can work on this
Please!