Add support for Microshift in CI

[vprashar2929]

Since we have added support for having microshift as a cluster in the local-dev-cluster repo and updated the kepler-action as well This PR is in continuation to add support for running Kepler integration tests on microshift cluster.

Dependency:
There is an open PR in Kepler-action to update some recent changes done to local-dev-cluster repo.

CC: @SamYuan1990 @rootfs

[rootfs]

@vprashar2929 new microshift integration test failed:

https://github.com/sustainable-computing-io/kepler/actions/runs/5437485041/jobs/9888072090?pr=759

+ for opt in ${DEPLOY_OPTIONS}
+ unset CI_DEPLOY
+ for opt in ${DEPLOY_OPTIONS}
+ unset OPENSHIFT_DEPLOY
+ echo 'Done ./hack/build-manifest.sh'
make[1]: Leaving directory '/home/runner/work/kepler/kepler'
Error response from daemon: No such container: microshift
make: *** [Makefile:263: cluster-sync] Error 1
Error: Process completed with exit code 2.

[vprashar2929]

Yes, currently it will fail as there is an open PR in Kepler-action for updating the logic. Once that is merged then CI will work fine

[SamYuan1990]

I will suggestion as bump up version in another PR.
As split this PR into 2 parts.

1. as adjust for logic, as add new verify.sh shell
2. bump up version to add microshift support.

[SamYuan1990]

@rootfs, @vprashar2929 for dev branch, we should keep this to main, for release we should use a release branch. Hence before we bump up version for kepler-action and local-dev-cluster, we should keep it as v0.0.0 and keep this PR open for review.

[SamYuan1990]

hence I will suggestion as bump up version in another PR.

[vprashar2929]

Sure bumping up the version in another PR sounds good. Till that time I will keep it as v0.0.0

[SamYuan1990]

we should considering change https://github.com/sustainable-computing-io/kepler/blob/main/manifests/config/exporter/patch/patch-ci.yaml

[SamYuan1990]

related with version bump up. Same comments here for version control.

[SamYuan1990]

to go with a new release or not for https://github.com/sustainable-computing-io/local-dev-cluster as k3s support
to go with a new release or not for sustainable-computing-io/kepler-action#50 as Prow support
@rootfs , @vprashar2929 , @jichenjc , @marceloamaral
I suppose we should have a vote as an agreement before we merge this PR.
and version bump up should also apply for helm and operator.
cc: @jiangphcn , @husky-parul

[vprashar2929]

One more thing I would like to add is when I was working on adding micro shift to CI I realized that building Kepler image takes more than 10GB of disk space(Dependent Images) on GH runner and we are using the same runner to create cluster, build image and deploy Kepler. Since the default disk size for GH runner is around 14GB. This make's difficult for the running cluster to run properly as I have seen the case where the cluster had to evict the running pods in order to reclaim the disk space because of NodeHasDiskPressure. To resolve this issue as of now I have added a step to stop the cluster build Kepler image push it to the local registry and then clean the images from the runner so that there is some disk space available and then restart the cluster. I am not a fan of this approach but I was thinking can we build the Kepler image on a separate runner upload it as an artifact and then download it on a different runner where we will launch our cluster and push it to the local registry. After that follow the normal process to deploy Kepler and trigger e2e tests?

@rootfs @SamYuan1990 @marceloamaral Any thoughts on this?

[rootfs]

@vprashar2929 that sounds a good idea. Would you like to have multiple jobs, kind/microshift/and potentially k3s in different jobs? How about we divide the flow into image build and image test? The image build job builds and uploads kepler image artifacts and the image test steps download the image and run kind/microshift/k3s in their runners.

[vprashar2929]

Yes, @rootfs I am also thinking of the same flow to have a job that builds and uploads Kepler image artifacts and another which uses our Kepler-action to bring up kind/microshift/k3s clusters, deploy Kepler, and run e2e test's on it.

[rootfs]

@SamYuan1990 is the request done?

[vprashar2929]

Incorporated review comments.🙂

[vprashar2929]

The current integration test will fail as I have reverted the bump up's in version for sustainable-computing-io/kepler-action. The v0.0.0 doesn't contain the functionality to launch microshift cluster.

[rootfs]

@vprashar2929 got a CI failure

[vprashar2929]

Yes @rootfs the CI will fail as we using v0.0.0 version of sustainable-computing-io/kepler-action that don't contain Microshift deployment. Post bump up the CI failures will stop. I have a draft PR ready to bump up versions. Waiting for the new version bump on kepler-action.