Add hack/build-dind-images.sh

[marun]

Building the dind images as part of a ci job is a frequent source of flakes caused by package installation failure via dnf. This change ensures the dind images are built by hack/build-base-images.sh so they can be baked into the ami.

Intended to help with #11452.

cc: @stevekuznetsov @openshift/networking

[stevekuznetsov]

Seems reasonable to me - this change will still result in them being built by the tests if they are not present, right? There will be a momentary disruption until we can get a new base_ami built.

[marun]

The images will still be built by the test runner. I'm counting on the test runner build being a no-op when the ami contains the correct built image, but if it changes it should rebuild.

[openshift-bot]

[Test]ing while waiting on the merge queue

[dcbw]

@marun any idea what the DNF failures are? I get them often locally too, and it seems like either or both of (a) systemctl restart docker or (b) removing the previous image eventually work. Can we get better logs out of the docker build?

[dcbw]

Jenkins failure is:

--> COPY openshift-generate-node-config.sh /usr/local/bin/
error: API error (500): {"message":"Error processing tar file(exit status 1): open /usr/local/bin/openshift-generate-node-config.sh: not a directory"}
[ERROR] PID 3189: hack/build-base-images.sh:11: `oc ex dockerbuild /data/src/github.com/openshift/origin/images/dind/node openshift/dind-node` exited with status 1.

[marun]

@dcbw afaict the dnf failures are mirror related and I don't know how local configuration could affect that.

Where are you seeing that jenkins failure? I would expect that kind of problem to be deterministic, and I can't replicate that locally. Are you able to reproduce?

[marun]

I'm assuming the failure around /usr/local/bin not being a directory was because /usr/local/bin did not exist and the first COPY in the dind Dockerfile didn't have a trailing slash. I've fixed that, let's see if it solves the problem.

[marun]

Updated to reflect @smarterclayton's recent changes.

[marun]

@stevekuznetsov looks good!

[smarterclayton]

I'm not in favor of building these on every build. Why do we need them?

[smarterclayton]

I'd be open to building these via a different script in the base ami.

[marun]

@smarterclayton: I'm not familiar with how the ami is built so I don't know how to accomplish what you're suggesting. Is it as simple as adding another script and modifying a jenkins job to target it?

[smarterclayton]

Probably. I'd just prefer to keep anything except the truly core images out of base since that's always built. hack/build-dind-images.sh is fine for now until we sort out what our image build strategy will be (we're going to scrap most of the current dockerfiles and replace them with package installed images eventually).

[marun]

Ok, adding a new script instead of modifying build-base-images.sh.

[stevekuznetsov]

@smarterclayton are you sure? AFAICT the build-base-images step is supposed to happen in AMI creation and build-images actually in the jobs. Is that no longer the case? If not, what's the point of build-base-images?

[smarterclayton]

build-base images is for the product. DIND is not part of the product. It's only for AMI and dev.

…

On Wed, Dec 14, 2016 at 9:02 AM, Steve Kuznetsov ***@***.***> wrote: @smarterclayton <https://github.com/smarterclayton> are you sure? AFAICT the build-base-images step is supposed to happen in AMI creation and build-images actually *in* the jobs. Is that no longer the case? If not, what's the point of build-base-images? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#12192 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABG_p9Tx-BcvQpFXAL42tfTUBefdvgbeks5rH_bvgaJpZM4LIDRY> .

[marun]

flake #12072 re-[test]

[openshift-bot]

Evaluated for origin test up to 365fce0

[openshift-bot]

continuous-integration/openshift-jenkins/test SUCCESS (https://ci.openshift.redhat.com/jenkins/job/test_pr_origin/12387/) (Base Commit: e9e7618)

[stevekuznetsov]

@marun You'll need to add a call of your script here to get the AMIs working with it.

[merge]

[openshift-bot]

Evaluated for origin merge up to 365fce0

[openshift-bot]

continuous-integration/openshift-jenkins/merge SUCCESS (https://ci.openshift.redhat.com/jenkins/job/test_pr_origin/12387/) (Base Commit: 039e433) (Image: devenv-rhel7_5549)