autodetect advertises address with IPv6

[aojea]

apiserver and kubelet advertise ipv6 endpoints when binding to ipv6 since 1.17

This removes the need to hardcode the kubelet and kube-apiserver IPs when using IPv6

Fixes : #1154

Unblocks kubernetes/kubernetes#86480

[liggitt]

/hold

it's unclear kubernetes/kubernetes#86480 is a compatible change... we should probably wait for a decision there before merging this

[aojea]

/hold

it's unclear kubernetes/kubernetes#86480 is a compatible change... we should probably wait for a decision there before merging this

yeah, thanks for the clarification, I've also added the WIP to this PR,
I'm finding IP families inconsistencies here too :/

[aojea]

/retest

[aojea]

2/2 failure of the same 2 tests, will investigate later
[sig-api-machinery] ResourceQuota should verify ResourceQuota with best effort scope. [Conformance]

[sig-api-machinery] AdmissionWebhook [Privileged:ClusterAdmin] should not be able to mutate or prevent deletion of webhook configuration objects [Conformance] expand_less

https://prow.k8s.io/view/gcs/kubernetes-jenkins/pr-logs/pull/sigs.k8s.io_kind/1195/pull-kind-conformance-parallel-ipv6/1208054793159839744/

/retest

[aojea]

It's failing because the kube-scheduler and the kube-controller-manager are being restarted constantly because they have the wrong http probes configured kubernetes/kubernetes#86493

[aojea]

🤔

https://storage.googleapis.com/kubernetes-jenkins/pr-logs/pull/sigs.k8s.io_kind/1195/pull-kind-conformance-parallel-ipv6/1213414923934109697/artifacts/logs/kind-control-plane/containers/kube-apiserver-kind-control-plane_kube-system_kube-apiserver-b5b1764ec4d32744472affd4653f94243e70947b6b5b888eb9a3b837e9035147.log

2020-01-04T11:13:47.951628264Z stderr F I0104 11:13:47.951353 1 server.go:596] external host was not specified, using 172.17.0.4

2020-01-04T11:13:48.948109526Z stderr F I0104 11:13:48.947888 1 rest.go:113] the default service ipfamily for this cluster is: IPv6

[aojea]

/hold cancel
all the changes needed were merged and we can benefit from the apiserver advertise address autodetection.
just only remains the kubelet to autodetect the node-ip family and we can get rid of hardcoding the node-address
kubernetes/kubernetes#85850
/assign @BenTheElder

[BenTheElder]

Is this going to break with older versions of kubernetes then...?

[aojea]

Is this going to break with older versions of kubernetes then...?

it will break only with kubeadm.k8s.io/v1beta2 , the commit didn't modify previous kubeadm templates so you can still deploy IPv6 clusters with previous kubernetes versions with kubeadm.k8s.io/v1beta1

Also, I'm not sure about this, but it may be possible to patch the preconfigured templates with a config file, no?

[aojea]

/hold
let's wait for
kubernetes/kubernetes#85850

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by: aojea
To complete the pull request process, please assign bentheelder
You can assign the PR to them by writing /assign @bentheelder in a comment when ready.

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

Needs approval from an approver in each of these files:

• OWNERS

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

[BenTheElder]

That's not good enough, we shouldn't break any released k8s.

This needs to be version gated.

[k8s-ci-robot]

@aojea: The following test failed, say /retest to rerun all failed tests:

Test name	Commit	Details	Rerun command
pull-kind-conformance-parallel-ipv6	4452848	link	/test pull-kind-conformance-parallel-ipv6

Full PR test history. Your PR dashboard. Please help us cut down on flakes by linking to an open issue when you hit one in your PR.

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository. I understand the commands that are listed here.

[aojea]

this is no autodetecting the node-ip using :: as parameter of kubelet because the node name doesn't resolve the IPv6 address, using the following code:

package main

import (
        "fmt"
        "net"
        "os"
        "runtime"
)

func main() {
        fmt.Println("GOOS:", runtime.GOOS)
        // Resolver
        addrs, _ := net.LookupIP(os.Args[1])
        fmt.Println("net.LookupIP addrs:", addrs)

}

we can confirm it

 /lookup kind-control-plane
GOOS: linux
net.LookupIP addrs: [172.17.0.2]

[aojea]

it was fixed in docker recently by this moby/libnetwork@e3bacd6

[aojea]

/hold
it seems we should wait for a new release of docker with the patch mentioned before included
If I understood docker releases correctly, it should be v19.03.6

[BenTheElder]

I don't think we can really rely on docker that new for our users?

…

On Sun, Feb 16, 2020 at 6:38 AM Antonio Ojea ***@***.***> wrote: /hold it seems we should wait for a new release of docker with the patch mentioned before included If I understood docker releases correctly, it should be v19.03.6 — You are receiving this because you were assigned. Reply to this email directly, view it on GitHub <#1195?email_source=notifications&email_token=AAHADK3RIW2WI6OJYOCOC4TRDFFVBA5CNFSM4J56DY72YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEL4IZPY#issuecomment-586714303>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAHADK7G4LVTMKZKIFQ6V5TRDFFVBANCNFSM4J56DY7Q> .

[aojea]

you are absolutely right ... I was so focused in this PR and chasing this odd behavior that forget these things 😅
/close

[k8s-ci-robot]

@aojea: Closed this PR.

In response to this:

you are absolutely right ... I was so focused in this PR and chasing this odd behavior that forget these things 😅
/close

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository.

[BenTheElder]

fixed in #1521