REQUEST: Copy kubernetes/dns to kubernetes-sigs/node-local-dns

[danwinship]

New repo, staging repo, or migrate existing

migrate existing

Is it a staging repo?

no

Requested name for new repository

node-local-dns

Which Organization should it reside

kubernetes-sigs

Who should have admin access?

aojea, bowei, danwinship, thockin

Who should have write access?

aojea, bowei, danwinship, thockin

Who should be listed as approvers in OWNERS?

keep existing OWNERS

Who should be listed in SECURITY_CONTACTS?

keep existing SECURITY_CONTACTS

What should the repo description be?

Kubernetes Node-Local DNS Cache

What SIG and subproject does this fall under?

sig-network kube-dns subproject

Please provide references to appropriate approval for this new repo

This is part of the plan for eventually retiring kubernetes/dns as discussed in February 26 SIG Network meeting and announced on the mailing list.

Part of the plan is that we need to migrate the still-actively-developed part of kubernetes/dns out to a new repo.

Additional context for request

The plan is to start with an exact copy of kubernetes/dns and then remove everything but the parts we want to keep. (Thus preserving history, etc.)

If you don't have a workflow for "copy repo" and it would be easier to do this as a straight move, we can just migrate https://github.com/danwinship/node-local-dns (a clean fork of kubernetes/dns) to https://github.com/kubernetes-sigs/node-local-dns.

cc @aojea @bowei @marqc

[aojea]

lgtm but I'd like to know if someone is actually stepping up to do the work

[danwinship]

I assume it will get as much work as it's getting now...

(I don't like the way the feature is designed and would rather kill it off, but I don't want to block deprecating kube-dns on being willing to deprecate node-local-dns too.)

[aojea]

@marqc does this impact you? I agree with dan it is better to have it in one independent home

[bowei]

Give the widespread usage of node local dns, I don't think we can kill it off without an alternative? I want to make it clear that the functionality is not in any way deprecated.

[marqc]

@marqc does this impact you? I agree with dan it is better to have it in one independent home

Yep, in GKE we still rely on NLD and it should not be deprecated. We can step up as new repo maintainers once I or my DNS team members get sig-network membership and as follow up clean up the code to remove everything unrelated to NLD.

[danwinship]

I'm not proposing to kill node-local-dns now. (The entire point of this PR is to not do that). I'm just saying we should eventually, because it's a bad and non-portable implementation of the functionality it's trying to provide. In particular, trafficDistribution: PreferSameNode gets us halfway to a better replacement.

I already have a PR ready to remove all the non-NLD bits of k/dns once we have the new repo.