Skip to content

Latest commit

 

History

History
 
 

kubefwd

Kubefwd

Author: Nick Santos

Kubernetes bulk service port-forwarding.

Want to use Kubernetes DNS addresses but don't want to run in a cluster?

kubefwd is a little server that updates your /etc/hosts so that addresses resolve like they would inside Kubernetes.

https://kubefwd.com/

Requirements

  • bash
  • kubefwd
  • entr
  • GNU core utils (tr, sort) - brew install coreutils

Examples

# kubefwd all namespaces Tilt deploys to.
v1alpha1.extension_repo(name='default', url='https://github.com/tilt-dev/tilt-extensions')
v1alpha1.extension(name='kubefwd:config', repo_name='default', repo_path='kubefwd')
# kubefwd all namespaces Tilt deploys to + manually configured additional namespace 'kubesystem'.
v1alpha1.extension_repo(name='default', url='https://github.com/tilt-dev/tilt-extensions')
v1alpha1.extension(name='kubefwd:config', repo_name='default', repo_path='kubefwd', args=['--namespaces=kubesystem'])

Usage

When Tilt starts up for the first time, it will prompt up your native OS GUI for your sudo password. Only kubefwd and its operator will get the sudo privileges.

The kubefwd operator will watch Tilt to see what namespaces you're deploying to, then configure kubefwd for that namespace.

If kubefwd gets wedged, Tilt has a custom button to the UI to refresh it. Click the kubefwd:run resource and press the button labelled "Refresh".

Future Work

Pod Discovery

If a pod inside the cluster restarts, you need to reboot kubefwd. You can do this manually from the tilt UI (the refresh button) or from the CLI with tilt trigger kubefwd:run. Could we do something more automatic?

Possible solutions:

  1. A system that watches the Tilt KubernetesDiscovery API and restarts it for you whenever the pods change.

  2. Even better, a version of kubefwd that reads the pods from the Tilt API.

Currently, this extension uses entr to restart kubefwd without requesting new credentials.

Multi-cluster kubefwd

Long-term, we'd like to be able to run kubefwd against multiple clusters at once (e.g., a read-only staging cluster and a writable dev cluster and a local KIND cluster.)