ahmetb / kubectl-foreach Goto Github PK

View Code? Open in Web Editor NEW

412.0 6.0 20.0 82 KB

Run kubectl commands in all/some contexts in parallel (similar to GNU xargs+parallel)

License: Apache License 2.0

Go 100.00%

kubectl kubectl-plugins

kubectl-foreach's Introduction

kubectl foreach

Run a kubectl command in one or more contexts (clusters) in parallel (similar to GNU parallel/xargs).

Usage

Usage:
    kubectl foreach [OPTIONS] [PATTERN]... -- [KUBECTL_ARGS...]

Patterns can be used to match context names from kubeconfig:
      (empty): matches all contexts
         NAME: matches context with exact name
    /PATTERN/: matches context with regular expression
        ^NAME: remove context with exact name from the matched results
   ^/PATTERN/: remove contexts matching the regular expression from the results
    
Options:
    -c=NUM     Limit parallel executions (default: 0, unlimited)
    -I=VAL     Replace VAL occurring in KUBECTL_ARGS with context name
    -q         Disable and accept confirmation prompts ($KUBECTL_FOREACH_DISABLE_PROMPTS) 
    -h/--help  Print help

Demo

Query a pod by label in minikube and *-prod* contexts:

$ kubectl foreach /-prod/ minikube -- get pods -n kube-system --selector compute.twitter.com/app=coredns --no-headers

     eu-prod | coredns-59bd9867bb-6rbx7   2/2     Running   0          78d
     eu-prod | coredns-59bd9867bb-9xczh   2/2     Running   0          78d
     eu-prod | coredns-59bd9867bb-fvn6t   2/2     Running   0          78d
    minikube | No resources found in kube-system namespace.
 useast-prod | coredns-6fd4bd9db4-7w9wv   2/2     Running   0          78d
 useast-prod | coredns-6fd4bd9db4-9pk8n   2/2     Running   0          78d
 useast-prod | coredns-6fd4bd9db4-xphr4   2/2     Running   0          78d
 uswest-prod | coredns-6f987df9bc-6fgc2   2/2     Running   0          78d
 uswest-prod | coredns-6f987df9bc-9gxvt   2/2     Running   0          78d
 uswest-prod | coredns-6f987df9bc-d88jk   2/2     Running   0          78d

Examples

Match to contexts by name: Run a command ("kubectl version") on contexts c1, c2 and c3:

kubectl foreach c1 c2 c3 -- version

Match to contexts by pattern: Run a command on contexts starting with gke (regular expression syntax):

kubectl foreach /^gke/ -- get pods

Match all contexts: empty context matches all contexts.

kubectl foreach -- version

Excluding contexts: Use the matching syntaxes with a ^ prefix to use them for exclusion. If no matching contexts are specified.

e.g. match all contexts except c1 and except those ending with prod (single quotes for escaping $ in the shell):

kubectl foreach ^c1 ^/prod'$'/ -- version

Using with kubectl plugins: Customize how context name is passed to the command (useful for kubectl plugins as --context must be specified after plugin name).

In this example, _ is replaced with the context name when calling "kubectl my_plugin".

kubectl foreach -I _ -- my_plugin -ctx=_

Limit parallelization: Only run 3 commands at a time:

kubectl foreach -c 3 /^gke-/

Install

Use Krew kubectl plugin manager:

kubectl krew install foreach

You can also build from source but you won't receive new version updates:

go install github.com/ahmetb/kubectl-foreach@latest

Remarks

Do not use this tool programmatically yet:

This tool is not intended for deploying workloads to clusters, or using programmatically. Therefore, it does not provide a structured output format or ordered printing that is meant to be parsed by or piped to other programs (maybe except for grep).

kubectl-foreach's People

Contributors

Stargazers

Watchers

kubectl-foreach's Issues

[Idea] Warn if --context=some-cluster is set

If -I is not set and --context=some-cluster is set, let the user know their command is about to just keep querying the same --context. Or maybe just fail.

swallowing output for non \n terminated lines

e.g.

...  -- get nodes -o=jsonpath='{.items[*].status.capacity.cpu}'`

doesn't print output that ends with a newline character, and nothing gets printed by the tool as a result.

The fix for this might be rather complicated, as prefixingWriter flushes on \n and currently there's no way to signal prefixingWriter to Flush(). (see Test_prefixingWriter https://github.com/ahmetb/kubectl-foreach/blob/4ac672dadc834895e723af04880d537df6b55af6/output_test.go#L43C6-L66).

[Idea] allow non-kubectl commands with -I

This can allow calling non-kubectl tools like

kubectl allctx -I _ [...] -- clusterctl --context=_ [...]

but it would complicate the UX notably. Maybe something like an -0 flag that indicates user must specify a command (and we'd print the command to be executed):

kubectl allctx -0 -I x -- kind delete clusters x

Question. Any option not to show "Will run command in context(s):"

Document exclusion with regexp in help/usage

 ^PATTERN: removes results from matched contexts

this makes people think ^/RE/ syntax is not an option.

Maybe call literals NAME and regexps /REGEXP/ or /PATTERN/.

error on 'go install'

when try to install

go install github.com/ahmetb/kubectl-allctx@latest

the output is

go: downloading github.com/ahmetb/kubectl-allctx v0.0.0-20220708173706-f2e07edffc5b
go: github.com/ahmetb/kubectl-allctx@latest: github.com/ahmetb/[email protected]: parsing go.mod:
	module declares its path as: kubectl-allctx
	        but was required as: github.com/ahmetb/kubectl-allctx

detect and fail when replacement character (-I) is not used

The other day I ran:

kubectl foreach -I _ [...bunch of filters...] -- get pods

but I didn't realize I was getting the same result from all clusters (because I didn't use _ anywhere, and -I caused --context= from not being passed to the kubectl cmd).

Fail if nothing is replaced among args when used with -I and tell the user they specified -I sequence but never used it.

[Idea] Generate structured output

If the cmd args contain -o=csv|yaml|json, maybe the tool can reuse that (or accept its own) to generate results to add context name and reformat the output.

[Idea] Allow configuration of pattern groups through a config file

What if there was a file that holds common queries, in a rc file somewhere on the home dir:

e.g.

context-groups:
  prod-new:
    include: "/prod/"
    exclude: "^/legacy/"
  stg:
    include: "/stg/"

$ kubectl allctx -g prod -- get nodes # g used as a flag for context group

[Idea] Being able to find clusters/contexts that have a matching k8s resource

So I forked this and added a feature that my team has been wanting, which is a command that allows you to find a context that contains a K8s resource we're looking for. This often comes up during oncall issues where a service or deployment is acting strange but we don't know where it's located.

Ex. kubectl foreach -- find svc foo -n bar

This would return any and all contexts that contain this resource. Not sure if this is something you'd be interested in for upstream or find useful, but figured I'll bring it up.

remove colors when piped to another program

|tee output still preserves colors (ideally must not).

[Bug] do not pass stdin

Prevent os.Exec() from inheriting the stdin so that commands like kubectl edit/exec fail, instead of hanging.

output table misalignment

im getting pods from multi clusters. result is looking like. is there bug or needed options or smth?

Some `Fprintf` statements aren't getting logged

I noticed that on my Apple M1, some of the Fprintf statements aren't getting properly logged and I'm not sure why. FWIW, I tested it in a Linux container and it worked fine. And when I changed the code to log using Fprintln, it worked fine.

For some reason Fprintf and Fprint don't log on certain statements unless a \n is added.

I've only really noticed this with the Continue? [Y/n] statement, but might be happening with other ones that I haven't noticed. I'm guessing it might be due to my M1. This happens on both Terminal and iTerm.

Adding a screen recording to demonstrate what I mean: https://recordit.co/H3j8XTSUMo