ci: rearchitect and streamline CI by indradhanush · Pull Request #143 · platform9/cluster-api-provider-bringyourownhost

indradhanush · 2026-07-20T12:29:39Z

With this PR, the GitHub CI will take ownership of building and pushing the agent bundle and controller-manager image directly to quay.io/platform9/cluster-api-provider-bringyourownhost. Additionally, we'll tag each artifact with its own unique tag derived from its git commit.

We also enable all the E2E suites to run in parallel, so some e2e specific changes were required here.

Note that: Now we run all E2E suites which was not the case previously. So E2E will now fail again from this PR onwards. To be fixed in follow ups.

Architecture

PR event / push to main / push tag with pattern `ci-*` (to explicitly trigger all CI runs)
  |
  |-- (unchanged, same as current behaviour) Security scan, golangci-lint,
  |    "Merge on lgtm label", Pull Request Labeler, markdown-links checks
  |
  |-- .github/workflows/build-agent-bundle.yml
  |     +-- build-host-agent-binary   (upload as artifact to reuse later in GitHub actions)
  |     |   
  |     +-- build-agent-bundle        (packages the .deb, upload as artifact to reuse later in GitHub actions; only push to quay for main)
  |
  |-- .github/workflows/unit-test.yml (standalalone, runs unit tests)
  |
  |-- .github/workflows/build-controller-manager.yml
  |     |
  |     +-- build-controller-manager   (upload as artifact to reuse later in GitHub actions; only push to quay for main)
  |
  \-- .github/workflows/e2e.yml
        |
        +-- e2e
             +-- poll for
             |   build-controller-manager and build-controller-manager jobs.
             |
             +-- download that job's image artifact
             |
             +-- run e2e tests concurrently with GINKGO_NODES=7

_{Stack created with GitHub Stacks CLI • Give Feedback 💬}

codecov · 2026-07-20T15:51:52Z

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

sebastian-pf9 · 2026-08-10T17:51:03Z

+  echo 'alias shasum="sha512sum"' >>~/.bashrc
+  # shellcheck disable=SC1090 # sourcing the user's own ~/.bashrc, not a repo file shellcheck can resolve
+  source ~/.bashrc


not super happy with this. this is pulling unknown stuff.

sebastian-pf9 · 2026-08-10T17:51:16Z

+  echo "removing build/ if already present"
+  rm -rf build/


make clean?

sebastian-pf9 · 2026-08-10T17:52:34Z

+IMAGE_REGISTRY_NAME_TAG=${IMAGE_REGISTRY}/${IMAGE_NAME_TAG}
+
+# make -C implicitly enables --print-directory on some GNU Make versions
+# (confirmed: not on this repo's dev-Mac Make 3.81, but yes on the Ubuntu


cut the noise

sebastian-pf9 · 2026-08-10T17:53:28Z

+
+  info "Verifying prerequisites"
+  #which aws > /dev/null || (echo "error: missing required command 'aws'" && exit 1)
+  which docker >/dev/null || (echo "error: missing required command 'docker'" && exit 1)


why bother? this will fail anyway later

sebastian-pf9 · 2026-08-10T17:59:01Z

+		bootstrapKubeconfigPath, err := uniqueTempFilePath("bootstrap-kubeconfig-*")
+		Expect(err).NotTo(HaveOccurred())
+		defer func() {
+			if removeErr := os.Remove(bootstrapKubeconfigPath); removeErr != nil {


bootstrap kubeconfig is broken iirc?

sebastian-pf9 · 2026-08-10T17:59:57Z

+		byoHostName1        = fmt.Sprintf("byohost1-%s", util.RandomString(6))
+		byoHostName2        = fmt.Sprintf("byohost2-%s", util.RandomString(6))


how can we verify we can re-use existing hostnames?

sebastian-pf9 · 2026-08-10T18:00:50Z

+		byoHostName1        = fmt.Sprintf("byohost1-%s", util.RandomString(6))
+		byoHostName2        = fmt.Sprintf("byohost2-%s", util.RandomString(6))


how can we verify we can re-use existing hostnames?

This needs to be a separate test where we test an explicit reuse.

sebastian-pf9 · 2026-08-10T18:01:12Z

 )

-var _ = Describe("When testing MachineDeployment scale out/in", func() {
+var _ = Describe("When testing MachineDeployment scale out/in [MD-Scale]", func() {


can we avoid running scale tests in the github action?

sebastian-pf9 · 2026-08-10T18:02:45Z

 	go run ./main.go

 docker-build: ## Build docker image with the manager.
+ifdef SKIP_BUILD


make was orignally built to handle build dependencies to avoid compiline and building files twice. This feels wrong

GINKGO_NODES runs multiple Ginkgo worker processes concurrently on the same runner, sharing one Docker daemon and filesystem (unlike a matrix-of-jobs design where each spec gets its own isolated VM). With hardcoded names like "byohost1" or "/tmp/host-agent1.log" shared across every spec file, two specs racing on the same runner would collide trying to create the same container or write the same log/kubeconfig path. Suffix host/container names and log file paths with a random string per spec, and generate the bootstrap kubeconfig's temp path via os.CreateTemp instead of a single hardcoded path, so concurrent nodes never step on each other's resources.

Split CI into 4 directly-triggered, standalone workflow files: - build-agent-bundle - build-controller-manager - unit-test - e2e The build workflows will only push from main. But otherwise, are pre-requisites for the E2E test, so that we do not rebuild the same artifact in E2E again, and instead can reuse it from the build's artifacts cache. We dont want to use GitHub's workflow_run, because it's not possible to show the status of the run on the PR / commit. As a result, use a standalone script to wait for the build workflows to complete.

github-actions Bot added the area/test-release label Jul 20, 2026

github-advanced-security Bot found potential problems Jul 20, 2026

View reviewed changes

Comment thread .github/workflows/build-agent-bundle.yml Fixed

Comment thread .github/workflows/build-agent-bundle.yml Fixed

Comment thread .github/workflows/build-controller-manager.yml Fixed

Comment thread .github/workflows/e2e.yml Fixed

indradhanush force-pushed the dg/ci-rearchitect-no-workflow-run branch from e92c0e5 to e5324bc Compare July 20, 2026 12:45

indradhanush force-pushed the dg/ci-rearchitect-no-workflow-run branch 6 times, most recently from deb13f1 to f2974bf Compare July 20, 2026 13:55

indradhanush changed the title ~~dg/ci rearchitect no workflow run~~ cI: rearchitect and streamline CI Jul 20, 2026

indradhanush changed the title ~~cI: rearchitect and streamline CI~~ ci: rearchitect and streamline CI Jul 20, 2026

indradhanush force-pushed the dg/ci-rearchitect-no-workflow-run branch from 509c444 to ffb44b9 Compare July 20, 2026 14:18

indradhanush marked this pull request as ready for review July 20, 2026 15:42

indradhanush force-pushed the dg/ci-rearchitect-no-workflow-run branch from ffb44b9 to 61db7a3 Compare July 20, 2026 17:38

indradhanush force-pushed the dg/ci-rearchitect-no-workflow-run branch from 61db7a3 to 61dd43e Compare July 31, 2026 16:01

indradhanush force-pushed the dg/ci-rearchitect-no-workflow-run branch from 61dd43e to 8c05983 Compare August 7, 2026 16:03

sebastian-pf9 approved these changes Aug 10, 2026

View reviewed changes

indradhanush force-pushed the dg/ci-rearchitect-no-workflow-run branch from 8c05983 to 84b0172 Compare August 11, 2026 13:31

Base automatically changed from dg/migrate-quay-registry to main August 11, 2026 13:58

indradhanush added 3 commits August 11, 2026 09:58

chore: Rename to match workflow file

24806b9

indradhanush force-pushed the dg/ci-rearchitect-no-workflow-run branch from 84b0172 to 24806b9 Compare August 11, 2026 13:58

indradhanush merged commit 0809a31 into main Aug 11, 2026
12 of 14 checks passed

sebastian-pf9 mentioned this pull request Aug 11, 2026

Some fixes to develop on Mac #161

Merged

		byoHostName1 = fmt.Sprintf("byohost1-%s", util.RandomString(6))
		byoHostName2 = fmt.Sprintf("byohost2-%s", util.RandomString(6))

Uh oh!

Conversation

indradhanush commented Jul 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Architecture

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

codecov Bot commented Jul 20, 2026

Codecov Report

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

indradhanush commented Jul 20, 2026 •

edited

Loading