loads kubeconfig only once

[dixudx]

What type of PR is this?

/kind bug
/sig cli
/area kubectl

What this PR does / why we need it:

Currently a single kubeconfig file will be loaded several times, which should be parsed for only once.

Which issue(s) this PR fixes (optional, in fixes #<issue number>(, fixes #<issue_number>, ...) format, will close the issue(s) when PR gets merged):
Fixes #69856

Special notes for your reviewer:
/cc @kubernetes/sig-cli-bugs
/assign soltysh

Does this PR introduce a user-facing change?:

kubectl loads config file once and uses persistent client config

[soltysh]

/lgtm
/approve

[soltysh]

/milestone v1.13
/priority important-longterm

[liggitt]

/hold

is this method expected to be self mutating? previously, it was non-mutating and theadsafe

[dixudx]

is this method expected to be self mutating? previously, it was non-mutating and theadsafe

it also freezes config at the point in time this method was first called. is that safe as well?

@liggitt @smarterclayton I don't think it is a good idea to load/parse the same config file again and again. That is time-consuming. And errors will occurr if we are changing the config file while the cli command is invoked.

This method is and will be only used by the command line. And only one such declaration is defined currently.

kubeConfigFlags := genericclioptions.NewConfigFlags()

I know your concern about thread safe. But kubectl only got one such declaration.

[liggitt]

@liggitt @smarterclayton I don't think it is a good idea to load/parse the same config file again and again. That is time-consuming. And errors will occurr if we are changing the config file while the cli command is invoked.

To be clear, I'm fine with kubectl updating to use a method that explicitly loads only once (or with kubectl updating to load once itself and propagate the config to the places that need it, rather than calling this method multiple times)

This method is and will be only used by the command line. And only one such declaration is defined currently.

This method is in an exported repo intended to be used by arbitrary command line tools. Searching only in the kubernetes/kubernetes repo does not indicate all usage.

[dixudx]

To be clear, I'm fine with kubectl updating to use a method that explicitly loads only once (or with kubectl updating to load once itself and propagate the config to the places that need it, rather than calling this method multiple times)

an acceptable alternative could be a second method that does something similar to this (though in a threadsafe manner) with a name that makes it clear the returned value is reusing a previously computed value if available, and switching callers that are ok with those semantics to that method

@liggitt Okay, I got it. I will have another alternative method to keep this peristent config.

This method is in an exported repo intended to be used by arbitrary command line tools. Searching only in the kubernetes/kubernetes repo does not indicate all usage.

Agree.

[dixudx]

@liggitt @smarterclayton I've added a flag for persistent client config and made it thread safe. And this change won't break current behavoirs.

For kubectl, we are switching using persistent client config.

[soltysh]

Two more nits and you'll be good to go.

[dixudx]

Kindly ping @soltysh to get this merged.

[soltysh]

/lgtm
/approve
Thanks for the patience

[soltysh]

/hold cancel

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: dixudx, soltysh

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• pkg/kubectl/OWNERS [soltysh]
• staging/src/k8s.io/cli-runtime/OWNERS [soltysh]
• staging/src/k8s.io/sample-cli-plugin/OWNERS [soltysh]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[fejta-bot]

/retest
This bot automatically retries jobs that failed/flaked on approved PRs (send feedback to fejta).

Review the full test history for this PR.

Silence the bot with an /lgtm cancel or /hold comment for consistent failures.

[dixudx]

/retest

[fejta-bot]

/retest
This bot automatically retries jobs that failed/flaked on approved PRs (send feedback to fejta).

Review the full test history for this PR.

Silence the bot with an /lgtm cancel or /hold comment for consistent failures.

[KnVerey]

Would it be possible to get this fix cherry-picked to a released version? I've been experiencing performance problems using kubectl with a large kubeconfig + large resource counts, so I would love to have this before 1.14 (It'd be great to see it in 1.11.x in fact).

[soltysh]

We prefer to cherry-pick only bug fixes, this is not quite that. It's more like a feature than a bug fix. Sorry.

[totahuanocotl]

@soltysh this issue did not exist in kubectl 1.10, and it has affected production systems where we rely on kubectl. I would call that a bug. We are using 1.11 because of the recent cve patch, so it would be good to have it in 1.11.