Reorganize vendor licenses again (revert #85220)

[thockin]

/kind cleanup

What this PR does / why we need it:

Try again to restructure license files for easier review. This is a revert and
cleanup on top of #85220, which was itself a revert of #76586.

NONE

[thockin]

/retest

[thockin]

/retest

[thockin]

/retest

[thockin]

@liggitt since you orchestrated the revert - any thoughts on how to get the failing signal to fail/pass BEFORE committing? Is there a way I can manually trigger those tests? @BenTheElder too

[liggitt]

paging back in what went wrong... I vaguely remember a different install path taken when testing from built CI artifacts (like we do in postsubmits) rather than building on the fly in the job itself (like we do in presubmits, maybe?)?

#85196 (comment)
looked like it pointed to
https://github.com/kubernetes/kubernetes/blob/master/cluster/gce/gci/configure.sh#L325

which looked like the output was missing a LICENSE file it previously had?

[thockin]

Yes, BUT dims@ tried to fix that and it failed differently. I am trying a different fix, but I need to iterate on kicking those tests off...

[BenTheElder]

I wasn't aware that CI vs presubmit was using a different install path on GCE ...
That rabbit hole may take some doing to explore ...

[BenTheElder]

cc @amwat @cheftako FYI re: GCE cluster scripts

[thockin]

Is there a way I can kick off the gci tests to see how they fare with this PR BEFORE merging?

[BenTheElder]

pull-kubernetes-e2e-gce* are running with kube-up on GCE (GCI is pretty meaningless at this point). there are a LOT of different things running on GCE with various options though. I don't think anybody knows all of them.

Kicking off periodic / postsubmit jobs from a PR is not something prow supports as far as I know, at best prow supports trying to convert from one type of job to another by hand in the config and having that be optionally triggered.

gce-cos-master-default should be the CI version of pull-kubernetes-e2e-gce I think.

[liggitt]

🎉

[liggitt]

changes lgtm after commit cleanup

[liggitt]

actually, maybe you already did the commit cleanup... in which case

/lgtm

[thockin]

For future me:

To test bazel building of the release:

mkdir -p $HOME/tmp/build_output
docker run -it \
  -e USER="$(id -u)" \
  -u="$(id -u)" \
  -v /etc/passwd.cache:/etc/passwd:ro \
  -v `pwd`:/src/workspace \
  -v $HOME/tmp/build_output:$HOME/tmp/build_output \
  -w /src/workspace \
  l.gcr.io/google/bazel:2.2.0 \
    --output_user_root=$HOME/tmp/build_output  \
    build \
    --sandbox_debug \
    --logging=6 \
    --verbose_failures \
    -- \
    //build/release-tars

Then you can look at the resulting tar files.

To test the GCI build:

rm -rf /tmp/workspace
mkdir -p /tmp/workspace/_artifacts
GOPATH=/tmp/go \
nice -n 19 \
kubetest \
  --dump=/tmp/workspace/_artifacts \
  --build --up --test --down \
  --provider=gce \
  --cluster=bootstrap-e2e \
  --gcp-network=bootstrap-e2e \
  --check-leaked-resources \
  --gcp-node-image=gci \
  --gcp-zone=us-west1-b \
  --runtime-config=api/all=true \
  --test_args=--ginkgo.focus=PodPreset.should.create.a.pod.preset \
  --ginkgo.skip= \
  --minStartupPods=8 \
  --build \
  --gcp-project=thockin-dev-2 \
  2>&1 | tee /tmp/workspace/build.log

This runs pretty fast since there is only on actual test running.

[thockin]

@liggitt or @dims - anyone who has to review deps -- this should make you happy, so I call upon you to review it. You probably can skim the commit that generates the files themselves

[liggitt]

tagged, drop the WIP prefix and it should be all set

[dims]

/approve
/lgtm

[justaugustus]

/retest

[thockin]

/retest

[thockin]

/retest

[thockin]

/retest

[BenTheElder]

this has gotten flakier :(
conntrack / TCP close_wait we have a test fix out for, just went back and reviewed that

[BenTheElder]

belated
/lgtm
/approve

bash / bazel / ... looks good 👍

Purge almost all memories of godep

unlikely 🙃

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: BenTheElder, dims, thockin

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

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [thockin]

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

[liggitt]

The statefulset flake is root caused and has a fix in the queue as well

[BenTheElder]

/retest
teardown / timeout failure for kubemark-e2e-gce-big ...

[fejta-bot]

/retest
This bot automatically retries jobs that failed/flaked on approved PRs (send feedback to fejta).

Review the full test history for this PR.

Silence the bot with an /lgtm cancel or /hold comment for consistent failures.

[thockin]

w00t. I did something useful. Kind of useful, anyway.

[dims]

Kudos Tim!