Skip to content

kubectl plugin - giving admins df (disk free) like utility for persistent volumes

License

Notifications You must be signed in to change notification settings

yashbhutwala/kubectl-df-pv

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

73 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

kubectl df-pv

A kubectl plugin to see df for persistent volumes.

Requirements

kube-apiserver has api/v1/nodes/ endpoint enabled

☑ Appropriate RBAC. This utility is meant for cluster-admin like user; specifically, you need a service account with enough RBAC privileges to access api/v1/nodes/ from the kube-apiserver.

☑ Using a storage provisioner that populates pv metrics in a compatible manner (see what's been tested below)

Quick Start

Installation

Via Krew

curl https://krew.sh/df-pv | bash
# . ~/.bashrc   # run if you use bash shell
# . ~/.zshrc    # run if you use zsh shell

From source

cd $GOPATH/src/github.com/
mkdir -p yashbhutwala
cd yashbhutwala/
git clone git@github.com:yashbhutwala/kubectl-df-pv.git
cd kubectl-df-pv/
make install
df-pv --help

Via Release Binary

macOS

download_path="./df-pv"
version="v0.2.2"
curl --fail -Lo $download_path "https://github.com/yashbhutwala/kubectl-df-pv/releases/download/${version}/kubectl-df-pv_${version}_darwin_amd64.tar.gz"
chmod +x $download_path
mv $(download_path) /some-dir-in-your-PATH/df-pv

Usage

kubectl df-pv

example output

Flags

> kubectl df-pv --help

df-pv emulates Unix style df for persistent volumes w/ ability to filter by namespace

It autoconverts all "sizes" to IEC values (see: https://en.wikipedia.org/wiki/Binary_prefix and https://kubernetes.io/docs/concepts/configuration/manage-resources-containers/#meaning-of-memory)

It colors the values based on "severity" [red: > 75% (too high); yellow: < 25% (too low); green: >= 25 and <= 75 (OK)]

Usage:
  df-pv [flags]

Flags:
  -h, --help               help for df-pv
  -n, --namespace string   if present, the namespace scope for this CLI request (default is all namespaces)
  -v, --verbosity string   log level; one of [info, debug, trace, warn, error, fatal, panic] (default "info")

Other useful commands

enable trace logging, but output to a file

df-pv -v trace 2> trace.log

Tested

Works on

GKE (kubernetes.io/gce-pd dynamic provisioner; both with ssd and standard)

kubeadm configured bare-metal cluster (rook ceph block dynamic provisioner using script)

Does not work due to storage provisioner

kind (rancher/local-path-provisioner dynamic provisioner)

minikube (gcr.io/k8s-minikube/storage-provisioner minikube-hostpath dynamic provisioner)

TODO

[ ] EKS

[ ] AKS

TODO Features

Yet to be completed

☒ sort-by flag

☒ exclude namespaces

☒ only show a specific colored result ("red", "yellow", "green")

Completed

df for all Persistent Volumes in the cluster

☑ human readable output as default (using IEC format)

☑ color based on usage [red: > 75% (too high); yellow: < 25% (too low); green: >= 25 and <= 75 (OK)]

☑ print PV name

☑ print volume mount name

Motivation

Have you ever wondered, "How much free disk space do all my PVs have?" Me too! That's why I built this plugin! I have always just wanted to quick way to see the disk usage of my Persistent Volumes (similar to df or du in Unix). It turns out I'm not the only one, there has been many upstream Kubernetes issues open again and again about this, even some KEPs and PRs. I have compiled some of the issues and KEPs that I've seen in the wild here:

Issues

"this feature is needed .. !!!!" - @halradaideh

"There was a plan to implement this for 1.7 but we ran out of time."

"I need to know to which Pod(s) a PVC is being mounted to, without having to go through each Pod one by one checking the Volume mounts and the PersistentVolumeClaim."

KEPs

"PVC should show how much of the available capacity is used vs available"

"Expose storage metrics to end users"

"Display PVC capacity usage using kubectl"... "Has this been continued somewhere? It would be super useful"

"exposing storage metrics to users"... "Status? Was it done?"

Other relevant/rabbit-hole links

"Volume metrics exposed in /stats/summary not available in /metrics"

interesting metric PR

something similar to du in metrics

client-go issue about kubelet api by @DirectXMan12

kubelet stats api

kubelet client

metrics-server client

blog about kubectl printers and columns

cli-runtime tableprinter usage

kubectl using cli-runtime; sorting is still not implemented