🐛 Check cluster is running for non-standalone Make targets

[bentito]

We have several non-standlone Make targets

• test-e2e
• kind-load
• install
• only-deploy-manifest

The above targets rely on Kind to be running.

Unfortunately, currently, errors returned from the Go code are pretty cryptic. Better would be an explicit catch and error when a Kind cluster is not found when these targets are run.

Currently we have:

catalogd checking_kind_running $ make test-e2e
/Users/btofel/go/bin/ginkgo-v2.20.2  -trace -vv  test/e2e
Running Suite: E2E Suite - /Users/btofel/workspace/catalogd/test/e2e
====================================================================
Random Seed: 1729265504

Will run 1 of 1 specs
------------------------------
[BeforeSuite]
/Users/btofel/workspace/catalogd/test/e2e/e2e_suite_test.go:33
  > Enter [BeforeSuite] TOP-LEVEL - /Users/btofel/workspace/catalogd/test/e2e/e2e_suite_test.go:33 @ 10/18/24 11:31:49.369

Ginkgo ran 1 suite in 4.681573708s

Test Suite Failed
make: *** [test-e2e] Error 1

wha?

but it would be better to have:

$ make test-e2e
if ! kubectl config current-context >/dev/null 2>&1; then \
		echo "Error: Could not get current Kubernetes context. Use 'run' or 'e2e' targets first."; \
		exit 1; \
	fi
Error: Could not get current Kubernetes context. Use 'run' or 'e2e' targets first.
make: *** [check-cluster] Error 1

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 38.23%. Comparing base (4d6327f) to head (2f186ef).
Report is 1 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #437   +/-   ##
=======================================
  Coverage   38.23%   38.23%           
=======================================
  Files          15       15           
  Lines        1224     1224           
=======================================
  Hits          468      468           
  Misses        706      706           
  Partials       50       50           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[joelanford]

This check might be a bit too presumptuous. For example, someone running test-e2e might want to run the tests after running kind create cluster --name=foobar and tilt up.

It may even be that someone wants to run the e2es against a non-kind cluster that already has OLM deployed.

[bentito]

Changed to make it more generic like this:

.PHONY: check-cluster
check-cluster:
	if ! kubectl config current-context >/dev/null 2>&1; then \
		echo "Error: Could not get current Kubernetes context. Use 'run' or 'e2e' targets first."; \
		exit 1; \
	fi

wdyt?

[bentito]

okay I changed the output to simply be:

echo "Error: Could not get current Kubernetes context. Maybe use 'run' or 'e2e' targets first?";

and check-cluster is called now from test-e2e, install and kind-load all of which really need some cluster running or they produce a non-descriptive and cryptic error message.

[camilamacedo86]

Hi @bentito

If you see the logs https://github.com/operator-framework/catalogd/actions/runs/11819448861/job/32929383546 (ci job of this PR)you will see that this check is wrong and is always failing:

I do not see the need for this check.
Instead, I would prefer the targets of this project to be as similar as possible to those of https://github.com/operator-framework/operator-controller so that life for us contributors is easier.

c/c @grokspawn