Re: [code-organization] Extracting hyperkube

[Stephen Augustus]

(+ SIG Release for vis)

On Fri, Aug 9, 2019, 18:22 Davanum Srinivas <dav...@gmail.com> wrote:

Team,

Do we know who is using hyperkube? We had an old issue where we were trying to transition everything to hyperkube[1], but i guess it feel out of favor and most folks just use the binaries. Right?

Can we get rid of hyperkube binary and image from release artifacts? 

Why am i asking this? when revisiting all the cloud provider dependencies that we drag in, it turns out we don't really need to ship cloud-controller-manager as we will have external providers that have their own controller manager. To remove cloud-controller-manager, we first have to remove it from hyperkube (deprecation in [2]). Then the question was, do we really need hyperkube?

Given that it's easy to just create a separate repo for hyperkube (prototype [3]). Do we have a set of folks who can sign up for this work? (Also there's k3s already! for the same use case).

Note that the only hiccup in this prototype was that we don't check in generated code for openapi which i have a fix we can debate on in [4].

WDYT? Can we give hyperkube images and binary the boot?

Thanks,

Dims

[1] https://github.com/kubernetes/kubernetes/issues/16508

[2] https://github.com/kubernetes/kubernetes/pull/81219

[3] https://github.com/dims/hyperkube

[4] https://github.com/kubernetes/kubernetes/pull/81239

--

Davanum Srinivas :: https://twitter.com/dims

--
You received this message because you are subscribed to the Google Groups "kubernetes-sig-architecture" group.
To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-arch...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/kubernetes-sig-architecture/CANw6fcE%3DCyzCTRqA6LY6u-Ypoj%3DQLOho3_Kbc6n-7Nks7iwFPA%40mail.gmail.com.

[Caleb Miles]

sounds good to me.

You received this message because you are subscribed to the Google Groups "kubernetes-sig-release" group.
To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-re...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/kubernetes-sig-release/CAFQm5yRtb4djUPDMSTFbXn3Fju3EKHvWgkL1coakyTZfbq%2BY1w%40mail.gmail.com.

[Mike Spreitzer]

hack/local-up-cluster.sh uses hyperkube, right?

Regards,
Mike

kubernetes-si...@googlegroups.com wrote on 08/09/2019 06:25:05 PM:

> From: Stephen Augustus <Ste...@agst.us>

> d/msgid/kubernetes-sig-architecture/
> CAFQm5yRtb4djUPDMSTFbXn3Fju3EKHvWgkL1coakyTZfbq%2BY1w%40mail.gmail.com.

[Davanum Srinivas]

yep. that could be easily modified to not use it.

[Davanum Srinivas]

Done! https://github.com/kubernetes/kubernetes/pull/81268

[Stephen Augustus]

(looks like the threads forked off, so adding SIG Release back)

I'm in favor of moving hyperkube out of tree.

-- Stephen

On Sun, Aug 11, 2019, 17:55 Davanum Srinivas <dav...@gmail.com> wrote:

Andrew,

Ack. If we move just hyperkube support to another repo, you can have custom Makefile/scripts for the musl use case as well :)

On Sun, Aug 11, 2019 at 2:53 PM Andrew Rynhard <and...@rynhard.io> wrote:

As Jason pointed out, Talos does indeed use hyperkube. That being said I have been looking for a good reason to get away from it. The main reason we do this is because we use musl instead of glibc and IIRC kubelet is linked against glibc.

> On Aug 11, 2019, at 11:49 AM, Lubomir I. Ivanov <neol...@gmail.com> wrote:
>
> AFAIK, SAP have claimed to use hyperkube in production.
> should / can we mark the image as deprecated in 1.16 and possibly
> remove in 1.17?
>
> lubomir
> --

>
> On Sat, 10 Aug 2019 at 01:22, Davanum Srinivas <dav...@gmail.com> wrote:
>>
>> Team,
>>
>> Do we know who is using hyperkube? We had an old issue where we were trying to transition everything to hyperkube[1], but i guess it feel out of favor and most folks just use the binaries. Right?
>>
>> Can we get rid of hyperkube binary and image from release artifacts?
>>

>> Why am i asking this? when revisiting all the cloud provider dependencies that we drag in, it turns out we don't really need to ship cloud-controller-manager as we will have external providers that have their own controller manager. To remove cloud-controller-manager, we first have to remove it from hyperkube (deprecation in [2]). Then the question was, do we really need hyperkube?

>>
>> Given that it's easy to just create a separate repo for hyperkube (prototype [3]). Do we have a set of folks who can sign up for this work? (Also there's k3s already! for the same use case).
>>
>> Note that the only hiccup in this prototype was that we don't check in generated code for openapi which i have a fix we can debate on in [4].
>>
>> WDYT? Can we give hyperkube images and binary the boot?
>>
>> Thanks,
>> Dims
>>
>> [1] https://github.com/kubernetes/kubernetes/issues/16508
>> [2] https://github.com/kubernetes/kubernetes/pull/81219
>> [3] https://github.com/dims/hyperkube
>> [4] https://github.com/kubernetes/kubernetes/pull/81239
>>
>>
>> --
>> Davanum Srinivas :: https://twitter.com/dims
>>
>> --
>> You received this message because you are subscribed to the Google Groups "kubernetes-sig-architecture" group.
>> To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-arch...@googlegroups.com.
>> To view this discussion on the web visit https://groups.google.com/d/msgid/kubernetes-sig-architecture/CANw6fcE%3DCyzCTRqA6LY6u-Ypoj%3DQLOho3_Kbc6n-7Nks7iwFPA%40mail.gmail.com.
>
> --

> You received this message because you are subscribed to the Google Groups "kubernetes-sig-cluster-lifecycle" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-cluster...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/kubernetes-sig-cluster-lifecycle/CAGDbWi_SqGqWk_GtusJbW6xn8sUbUaUXah%2BNg%2BEOXNC9AXdHtw%40mail.gmail.com.

--

Davanum Srinivas :: https://twitter.com/dims

--
You received this message because you are subscribed to the Google Groups "kubernetes-sig-architecture" group.
To unsubscribe from this group and stop receiving emails from it, send an email to kubernetes-sig-arch...@googlegroups.com.

To view this discussion on the web visit https://groups.google.com/d/msgid/kubernetes-sig-architecture/CANw6fcE1hsOgkRtNeFws3s65ZTRvFZonLRkTmfdd4c59hnVdRA%40mail.gmail.com.

[David Eads]

I like the idea of moving away from hyperkube.  I'm less sure that an out of tree hyperkube is practical long term if the mono-repo is someday divided.

To view this discussion on the web visit https://groups.google.com/d/msgid/kubernetes-sig-architecture/CAFQm5ySX1pDOni2usf8Y--Po2o94VvK-5R9PaCGptAozc0FLwA%40mail.gmail.com.

[Patrick Lang]

Is your goal to do this during 1.16 or 1.17?

 

The test passes running on Azure depend on this and build the hyperkube image from kubetest. Moving away from this would need a week or two so we can coordinate across the teams using it since they span US/China/EU. I think that’s too risky this close to feature freeze.

 

https://github.com/kubernetes/test-infra/blob/ee5d952705281807c0528babdffde85c0e9290d8/kubetest/azure.go#L628

To view this discussion on the web visit https://groups.google.com/d/msgid/kubernetes-sig-architecture/CANw6fcHtGFwCt_cH5MZP9xv9S7mRqjQR8tDpb4K3j%3DJgM9LcWQ%40mail.gmail.com.

[Davanum Srinivas]

Patrick,

1.17 :) So no worries!

[Stephen Augustus]

SIG Arch Chairs --

Can you approve this migration request: https://github.com/kubernetes/org/issues/1365

-- Stephen

On Sun, Aug 11, 2019 at 5:57 PM Stephen Augustus <Ste...@agst.us> wrote: