Release blocking issues (1.19.0-beta.0, 1.18.3, 1.17.6, 1.16.10)

[justaugustus]

There are a few issues blocking releases right now and I realized there's not a singular place tracking them, so I'm opening this up.

Required

kube-cross and k8s-cloud-builder:

• (kube-cross: Support building against multiple Go versions in parallel #1258) kube-cross: Support building against multiple Go versions in parallel
• (Revert k8s-cloud-builder changes in kube-cross and enable variant building #1279) Revert k8s-cloud-builder changes in kube-cross and enable variant building
• (k8s-cloud-builder: Build images on kube-cross v1.14.2-4 and v1.13.9-5 #1280) k8s-cloud-builder: Build images on kube-cross v1.14.2-4 and v1.13.9-5
• Update to kube-cross:v1.13.9-5:
  • master: base-images: Update to kube-cross:v1.13.9-5 kubernetes#90963
  • release-1.18: [1.18] base-images: Update to kube-cross:v1.13.9-5 kubernetes#90964
  • release-1.17: [1.17] base-images: Update to kube-cross:v1.13.9-5 kubernetes#90965
  • release-1.16: [1.16] base-images: Update to kube-cross:v1.13.9-5 kubernetes#90966

Branch cut:

• (Allow gcbmgr and anago to build betas from master #1286, lib/release: Fix version label logic for release cuts #1288) [anago] Support cutting branches at RC1 instead of beta.0 - Change release branching strategy to create release-1.x branch late #1207
• ([krel] gcbmgr: Add validation method for GcbmgrOptions #1293) [krel] gcbmgr: Add validation method for GcbmgrOptions

Release notes:

• Release notes on k/k fail to generate w/ new dependencies feature - Release notes on k/k fail to generate w/ new dependencies feature #1295

Nice-to-have

• Base image updates:
  • go-runner (1.19-only): New go-runner image for distroless scenarios kubernetes#90804, Switch core master base images (kube-apiserver, kube-scheduler) from debian to distroless kubernetes#90674
  • debian-base / debian-iptables:
    • master: cleanup: Base image workflow improvements kubernetes#90665, base-images: Use new debian-base:v2.1.0 and build new debian-iptables:v12.1.0 kubernetes#90697, and base-images: Use new debian-iptables:v12.1.0 kubernetes#90782
    • release-1.18: [1.18] base-images: Use debian-base:v2.1.0 and debian-iptables:v12.1.0 (includes CVE fixes) kubernetes#90863
    • release-1.17: [1.17] base-images: Use debian-base:v2.1.0 and debian-iptables:v12.1.0 (includes CVE fixes) kubernetes#90938
    • release-1.16: [1.16] base-images: Use debian-base:v2.1.0 and debian-iptables:v12.1.0 (includes CVE fixes) kubernetes#90940
• Mismatched or dirty local k/release copy does not block krel gcbmgr runs - Mismatched or dirty local k/release copy does not block krel gcbmgr runs #1278

/assign
/kind bug
/priority critical-urgent
cc: @kubernetes/release-engineering

[justaugustus]

/priority critical-urgent

[saschagrunert]

Proposal for the branch cutting changes:

1. Add a new variant beta to the current krel gcbmgr --type flag, which is currently:
   --type string Release type (must be one of: 'prerelease', 'rc', 'official')
2. Pass that flag via the GCB substitutions to anago
3. Modify anago (this seems to be the funny part) to not enforce alpha releases on the master and allow betas by taking the new release type into account. I guess one (of some) critical code part(s) is here:

   release/anago

   Lines 471 to 492 in c600d12

   	# NOTE: For new branches, files are still also updated on master (CHANGELOG).
   	# Therefore, while the new branch may be based on a commit earlier than
   	# HEAD, all master activity must occur AT HEAD.
   	if [[ -n "$PARENT_BRANCH" ]]; then
   	if [[ $RELEASE_VERSION_PRIME == $version ]]; then
   	if [[ $tree_object =~ release- ]] && \
   	! git rev-parse $tree_object >/dev/null 2>&1; then
   	# Only create/reset (-B) and set a branch_point if the *release-*
   	# branch doesn't already exist locally
   	branch_arg="-b"
   	# Use BRANCH_POINT if set, otherwise, the hash from
   	# JENKINS_BUILD_VERSION
   	branch_point=${BRANCH_POINT:-$commit}
   	fi
   	else
   	# if this is not the PRIMary version on the named RELEASE_BRANCH, use the
   	# parent
   	tree_object=$PARENT_BRANCH
   	fi
   	else
   	[[ $label == "alpha" ]] && tree_object="$commit"
   	fi

[justaugustus]

Proposal for the branch cutting changes:

@saschagrunert -- Let's bring this discussion to the branch strategy issue. Commented here: #1207 (comment)

[justaugustus]

This is complete.
/close

[k8s-ci-robot]

@justaugustus: Closing this issue.

In response to this:

This is complete.
/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.