Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/nearmap/kubefs

User friendly kubernetes tools to make your clusters less mysterious and more accessible
https://github.com/nearmap/kubefs

filesystem fuse hackday k8s kubernetes

Last synced: about 2 months ago
JSON representation

User friendly kubernetes tools to make your clusters less mysterious and more accessible

Awesome Lists containing this project

README 

        
# User friendly kubernetes tools



This project is a collection of tools created to demystify your kubernetes

clusters and make them more accessible to non-expert users.



* `kubefs` is a filesystem that allows you to browse the kubernetes objects in

  your cluster in a familiar way, as files and directories.

* `podview` is a terminal program which gives you a real time view of pods that

  interest you, across clusters. You can use it to watch code deployments in

  real time, or check up on the health of your workloads.



## kubefs - a fuse filesystem for browsing clusters



`kubefs` is a **read-only** filesystem that runs in user space (you don't need

to be `root` to mount it) that allows you to browse objects in your Kubernetes

clusters.



![kubefs screenshot](docs/assets/kubefs-shot.png)



It loads your kube config(s) from `$KUBECONFIG` or `~/.kube` and uses that to

present a top level view for you to navigate. The modification timestamps of the

files and directories reflect the `creationTimestamp` of the kube objects

themselves.



Behind the scenes, `kubefs` makes requests to the k8s API server to fetch all

these objects and populate the filesystem. This can be slow, so directory

entries are cached.



## podview - a real time dashboard for pods



`podview` is a curses based terminal program which gives you a real time view of the

pods that you want to see.



It loads your kube config(s) from `$KUBECONFIG` or `~/.kube` and uses that to

detect all your clusters. You can then filter on:



* cluster name using `-c` / `--cluster`

* namespace name using `-n` / `--namespace`

* pod name using `-p` / `--pod`



A very common case is watching the state of pods for a particular

workload/service across all your clusters:



```bash

$ ./pv --pod 'prom*'

```



![podview screenshot](docs/assets/podview-shot.png)



On startup `podview` first lists all the pods matching your filter, and then

proceeds to watch them for updates.



Pods are listed per cluster, and sorted by `creationTimestamp` so you will see

the oldest pods at the top.



Keyboard controls:



* Arrow keys to scroll horizontally or vertically by one character.

* PageUp/PageDown to scroll vertically by half a page.



## Getting started



Requirements:

​

* Python 3.8 or later.

* A working kube config (make sure `kubectl` works).

* Additional dependencies for `kubefs` (not required for `podview`):

  * `fuse` (available on Linux and [Mac](https://osxfuse.github.io/))

  * Somewhere to mount the filesystem. `~/kubeview` is the suggested mount

    point, but you can use any location you like.



### Ubuntu



#### System packages



Let's first check the version of your system Python:



```bash

$ python3 -V

Python 3.8.10

```



Install the fuse packages and the python-venv package matching your installed

version:



```bash

$ sudo apt install fuse libfuse2 python3.8-venv

```



### Project setup (the scripted way)



Clone the repository:



```bash

$ git clone https://github.com/nearmap/kubefs

$ cd kubefs

```



These scripts automate the manual setup below:



```bash

# kubefs

$ mkdir -p ~/kubeview

$ ./kfs ~/kubeview



# podview

$ ./pv

```



They basically manage the virtual environment for you, so *they have to be run

outside of the virtual environment*.



The scripts assume that you either:

- Don't have a virtual environment in `.ve/` at all (it will setup it up for

  you), or

- You have a fully populated virtual environment in `.ve/` (it will assume it's

  up to date to avoid running `pip install` every time).



If your virtual environment isn't working try removing it by doing `rm -rf .ve`

and re-running the scripts. See also the `Troubleshooting` section below.



### Project setup (the manual way)



Clone the repository:



```bash

$ git clone https://github.com/nearmap/kubefs

$ cd kubefs

```



Create the virtual environment by doing:



```bash

$ python3 -m venv .ve

```



Activate it by doing:



```bash

$ . .ve/bin/activate

(.ve) $

```



Once you've activated the virtual environment install the dependencies into it:



```bash

(.ve) $ pip install -r requirements.txt

```



Finally, make sure kubefs and podview can be started without errors:



```bash

# kubefs

(.ve) $ mkdir -p ~/kubeview

(.ve) $ bin/kubefs ~/kubeview



# podview

(.ve) $ bin/podview

```



### Troubleshooting



**When running `kubefs` or `podview` I get `ModuleNotFoundError: No module named

'fuse'`**



Your virtual environment is missing one or more dependencies. Try re-creating

it and re-installing the dependencies. See the `Project setup` steps above.



**While creating/installing dependencies into a virtual environment I see

errors in red, something about `Failed building wheel`.**



These are not fatal errors and you can ignore them.



**When running `kubefs` or `podview` I get `FileNotFoundError: [Error 2] No such

file or directory: '/home/user/.kube'`.**



Your kube config files could not be detected. (Does `kubectl` work?)



If you have `$KUBECONFIG` set make sure it's pointing at one or more valid kube

config files, eg.

`KUBECONFIG=/home/user/.kube/cluster1:/home/user/.kube/cluster2`.



Otherwise, make sure that your `~/.kube` directory contains at least one valid

kube config file.



**Using logs to troubleshoot problems that occur at runtime.**



`kubefs` runs in the foreground and logs to stdout/stderr.



`podview` runs a curses ui in the foreground. It logs to `var/log/podview.log`.