Can't replicate `kubectl apply -f` functionality with Python client

[Datamance]

What happened (please include outputs or screenshots):
The kubectl CLI uses a StrategicMergePatch type in the underlying API request for apply.

The Python client should be able to do the same by letting us pass a Content-Type of 'application/strategic-merge-patch+json' into the headers dict, but this odd little piece of generated code precludes us from doing so

python/kubernetes/client/api/apps_v1_api.py

Line 3889 in ccd3ce4

header_params['Content-Type'] = self.api_client.select_header_content_type( # noqa: E501

I would like to leverage the Python client as much as possible without resorting to kubectl. Is there a way to get around this for the time being?

Thanks in advance!

Environment:

• Kubernetes version (kubectl version):

• OS (e.g., MacOS 10.13.6):
  10.15.1

• Python version (python --version)
  3.8.0

• Python client version (pip list | grep kubernetes)
  10.0.1

[Datamance]

OK, I see there appears to be some activity on this precise issue now. The Java client seemed to be having the same kind of issue in 2017, to boot. Codegen seems to be a bit unwieldy.

Still interested in any workarounds that stay in Python-land!

[roycaihw]

We patched the generated rest client to allow using strategic merge patch:

python/kubernetes/client/rest.py

Lines 158 to 161 in ccd3ce4

	if headers['Content-Type'] == 'application/json-patch+json':
	if not isinstance(body, list):
	headers['Content-Type'] = \
	'application/strategic-merge-patch+json'

That being said, kubectl apply is more than sending a strategic merge patch-- it has logic to do a three-way diff to compose the patch. A better way to do that is to use the new server-side apply feature kubernetes/enhancements#555. We should support sending patch with application/apply-patch content type

[Datamance]

@roycaihw it looks like this feature is not generally available as of 1.17?

[roycaihw]

No it is not. From what I heard it was supported for a sub-portion of Kubernetes APIs in 1.15.

[Datamance]

@roycaihw given that this client doesn't implement the three-way diff logic found in the kubectl tool, and you don't yet support hooking into the server-side apply feature, if we want to emulate this behavior right now - do we have to shell out to kubectl? If so, when would you expect the server-side apply feature to be generally available, and how long after that would you be adding support in the Python client?

Related - is this the logic that you're referring to? If so, I might try my hand at implementing it myself - especially if the only two requirements are 1) getting the patch body correct and 2) having the correct content-type.

[Datamance]

also @roycaihw doesn’t that rest client patch preclude me from doing anything other than a strategic merge patch? I’m on a mobile device so it’s a little hard to look at code right now but I’m unclear as to when body is going to be a list - perhaps this is only then for when I’m patching multiple resources? But then, what if I want to do a strategic merge patch to multiple resources? Or do another type of patch on a single resource?

[couaky]

Hi !

As I understand with the current patched generator, all PATCH requests will use application/strategic-merge-patch+json or application/json-patch+json if body is a list and there is no workaround to use the other type application/merge-patch+json until the mentioned merge request will be done ?

Thanks.

[Datamance]

@roycaihw when approximately will this client be supporting the application/apply-patch content type? Also I'm guessing that they'll support a +json prefix as well but just wanted to make 100% sure given that the documentation only mentions a application/apply-patch+yaml

[Datamance]

@roycaihw and @micw523 - it looks like the two of you are the major contributors to this repository. Are you able to answer to some of the questions above, or at least provide some sort of status update?

[micw523]

I'd be watching the status of PR #959. There are propositions of how to deal with it, but it might require upstream (generator) changes.

[fejta-bot]

Issues go stale after 90d of inactivity.
Mark the issue as fresh with /remove-lifecycle stale.
Stale issues rot after an additional 30d of inactivity and eventually close.

If this issue is safe to close now please do so with /close.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/lifecycle stale

[fejta-bot]

Stale issues rot after 30d of inactivity.
Mark the issue as fresh with /remove-lifecycle rotten.
Rotten issues close after an additional 30d of inactivity.

If this issue is safe to close now please do so with /close.

Send feedback to sig-testing, kubernetes/test-infra and/or fejta.
/lifecycle rotten