GNU bug report logs - #59797
30.0.50; [wishlist] Using namespaces in Tramp's kubernetes integration

Previous Next

Package: emacs;

Reported by: Michael Albinus <michael.albinus <at> gmx.de>

Date: Sat, 3 Dec 2022 09:21:01 UTC

Severity: wishlist

Found in version 30.0.50

Fixed in version 31.1

Done: Michael Albinus <michael.albinus <at> gmx.de>

Bug is archived. No further changes may be made.

Full log


Message #41 received at 59797 <at> debbugs.gnu.org (full text, mbox):

From: Eshel Yaron <me <at> eshelyaron.com>
To: Michael Albinus <michael.albinus <at> gmx.de>
Cc: Aaron Gonzales <aarongonzales1 <at> gmail.com>, 59797 <at> debbugs.gnu.org,
 Filipp Gunbin <fgunbin <at> fastmail.fm>, Warren Lynn <wrn.lynn <at> gmail.com>
Subject: Re: bug#59797: 30.0.50; [wishlist] Using namespaces in Tramp's
 kubernetes integration
Date: Tue, 23 Jul 2024 14:59:52 +0200
Michael Albinus <michael.albinus <at> gmx.de> writes:

> Eshel Yaron <me <at> eshelyaron.com> writes:
>
>> Hi Michael,
>
> Hi Eshel,
>
>> Michael Albinus <michael.albinus <at> gmx.de> writes:
>>
>>> Now we have kubernetes integrated in Tramp. We can access pods like
>>> "/kubernetes:POD:/path/to/file".
>>>
>>> However, kubernetes can much more. For all services (like pods) it
>>> creates DNS entries "<service-name>.<namespace-name>.svc.cluster.local".
>>> Should'nt we support this? It means should we allow to access kubernetes
>>> pods like "/kubernetes:POD.NAMESPACE:/path/to/file", including
>>> completion for namespace names?
>>
>> [...]
>>
>>> Finally, I've taken the stab and implemented something along these
>>> lines. It is a little bit different from earlier proposals.
>>>
>>> - You can now specify a container in the host name, dotted notation.
>>>   "/kubernetes:container.pod:" accesses the given container. If it is
>>>   not specified, as in "/kubernetes:pod:", the first container is taken.
>>>
>>> - You can now specify different context and namespace. Change the new
>>>   user options tramp-kubernetes-context or tramp-kubernetes-namespace
>>>   for this effect.
>>>
>>> Everything pushed to the Emacs and Tramp master repositories. The target
>>> is now Tramp 2.7 / Emacs 30. Comments welcome :-)
>>
>> Sorry for not chiming in when this discussion was still ongoing, and
>> thank you for your work on the Kubernetes integration.  I use it a lot.
>>
>> Would it be possible the reconsider adding an optional namespace to
>> Tramp's Kubernetes file name syntax?  I'm thinking of something like
>> "/kubernetes:[CONTAINER.]POD[.NAMESPACE]:/...".
>
> Possible. However, the syntax is ambiguous. If we have "/kubernetes:name1.name2:/..."
> it isn't clear what they are. name1 could be a container, and name2 a pod. Or
> name1 is a pod, and name2 is a namespace. Could we agree on another
> delimiter between POD and NAMESPACE?

You're absolutely right.  AFAIU from the Kubernetes documentation[0],
non-alphanumeric characters other than "-" and "." should be safe.
Maybe "@" would be the most natural choice?  "+" could work too I think.

>> My use case is that I want to start Dired for a (container in a) pod in
>> another namespace.  I can almost get there if I temporarily bind
>> tramp-kubernetes-namespace around the call to dired, but the binding
>> doesn't carry on to subsequent operations in the Dired buffer.  So an
>> explicit namespace in default-directory could help, I think.
>
> I see, it makes sense. My practical kubernetes experience is sparse; I
> would exploit you while implementing. At least for tests.

SGTM.

> Note also, that Emacs 30 is in pretest. Any such feature change will be
> targeted to Emacs 31 / Tramp 2.8.

Of course, that's fine.

Thank you!


[0] https://kubernetes.io/docs/concepts/overview/working-with-objects/namespaces/
    https://kubernetes.io/docs/concepts/overview/working-with-objects/names/




This bug report was last modified 303 days ago.

Previous Next


GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.