ci: follow-up changes and fix flaky Windows tests. Fixes #12744 (#13670)

Signed-off-by: Mason Malone <mmalone@adobe.com>

.github/workflows/ci-build.yaml

@@ -240,10 +240,11 @@ jobs:
             profile: minimal
             use-api: false
     steps:
-      - name: Free up disk space
+      - name: Free up unused disk space
         run: |
           printf "==> Available space before cleanup\n"
           df -h
+          # these directories are not used by E2E tests
           sudo rm -rf /usr/share/dotnet /usr/local/lib/android /opt/ghc /usr/local/.ghcup /opt/hostedtoolcache/CodeQL
           printf "==> Available space after cleanup\n"
           df -h
@@ -275,14 +276,10 @@ jobs:
             export INSTALL_K3S_VERSION=${{ matrix.install_k3s_version }}
           fi
 
-          cat >/tmp/kubelet.config <<EOT
-          apiVersion: kubelet.config.k8s.io/v1beta1
-          kind: KubeletConfiguration
-          imageMinimumGCAge: 1h
-          imageGCHighThresholdPercent: 100
-          EOT
-
-          curl -sfL https://get.k3s.io | INSTALL_K3S_CHANNEL=stable INSTALL_K3S_EXEC='--docker --kubelet-arg=config=/tmp/kubelet.config' K3S_KUBECONFIG_MODE=644 sh -
+          curl -sfL https://get.k3s.io | INSTALL_K3S_CHANNEL=stable \
+            INSTALL_K3S_EXEC="--docker --kubelet-arg=config=${GITHUB_WORKSPACE}/test/e2e/manifests/kubelet-configuration.yaml" \
+            K3S_KUBECONFIG_MODE=644 \
+            sh -
           until kubectl --kubeconfig=/etc/rancher/k3s/k3s.yaml cluster-info ; do sleep 10s ; done
           cp /etc/rancher/k3s/k3s.yaml /home/runner/.kubeconfig
           echo "- name: fake_token_user" >> $KUBECONFIG

test/e2e/manifests/kubelet-configuration.yaml

@@ -0,0 +1,11 @@
+# The E2E CI workflow passes this file to k3s using "--kubelet-arg=config=".
+# Docs: https://docs.k3s.io/cli/agent?_highlight=kubelet&_highlight=arg#customized-flags
+
+# API reference: https://kubernetes.io/docs/reference/config-api/kubelet-config.v1beta1/
+apiVersion: kubelet.config.k8s.io/v1beta1
+kind: KubeletConfiguration
+# Prevent images from being GC'd during a test run (which will lead to ErrImageNeverPull errors) by
+# raising the minimum age to 1 hour and the disk usage threshold to 100%.
+# We probably don't need to specify both, but it doesn't hurt.
+imageMinimumGCAge: 1h
+imageGCHighThresholdPercent: 100

workflow/artifacts/http/http_test.go

@@ -5,7 +5,7 @@ import (
 	"net/http"
 	"net/http/httptest"
 	"os"
-	"path"
+	"path/filepath"
 	"testing"
 
 	"github.com/stretchr/testify/assert"
@@ -20,32 +20,37 @@ func TestHTTPArtifactDriver_Load(t *testing.T) {
 	a := &wfv1.HTTPArtifact{
 		URL: "https://github.com/argoproj/argo-workflows",
 	}
+	tempDir := t.TempDir()
+
 	t.Run("Found", func(t *testing.T) {
+		tempFile := filepath.Join(tempDir, "found")
 		err := driver.Load(&wfv1.Artifact{
 			ArtifactLocation: wfv1.ArtifactLocation{HTTP: a},
-		}, "/tmp/found")
+		}, tempFile)
 		require.NoError(t, err)
-		_, err = os.Stat("/tmp/found")
+		_, err = os.Stat(tempFile)
 		require.NoError(t, err)
 	})
 	t.Run("FoundWithRequestHeaders", func(t *testing.T) {
+		tempFile := filepath.Join(tempDir, "found-with-request-headers")
 		h1 := wfv1.Header{Name: "Accept", Value: "application/json"}
 		h2 := wfv1.Header{Name: "Authorization", Value: "Bearer foo-bar"}
 		a.Headers = []wfv1.Header{h1, h2}
 		err := driver.Load(&wfv1.Artifact{
 			ArtifactLocation: wfv1.ArtifactLocation{HTTP: a},
-		}, "/tmp/found-with-request-headers")
+		}, tempFile)
 		require.NoError(t, err)
-		_, err = os.Stat("/tmp/found-with-request-headers")
+		_, err = os.Stat(tempFile)
 		require.NoError(t, err)
-		assert.FileExists(t, "/tmp/found-with-request-headers")
+		assert.FileExists(t, tempFile)
 	})
 	t.Run("NotFound", func(t *testing.T) {
+		tempFile := filepath.Join(tempDir, "not-found")
 		err := driver.Load(&wfv1.Artifact{
 			ArtifactLocation: wfv1.ArtifactLocation{
 				HTTP: &wfv1.HTTPArtifact{URL: "https://github.com/argoproj/argo-workflows/not-found"},
 			},
-		}, "/tmp/not-found")
+		}, tempFile)
 		require.Error(t, err)
 		argoError, ok := err.(errors.ArgoError)
 		require.True(t, ok)
@@ -55,25 +60,29 @@ func TestHTTPArtifactDriver_Load(t *testing.T) {
 
 func TestArtifactoryArtifactDriver_Load(t *testing.T) {
 	driver := &ArtifactDriver{Client: http.DefaultClient}
+	tempDir := t.TempDir()
+
 	t.Run("NotFound", func(t *testing.T) {
+		tempFile := filepath.Join(tempDir, "not-found")
 		err := driver.Load(&wfv1.Artifact{
 			ArtifactLocation: wfv1.ArtifactLocation{
 				Artifactory: &wfv1.ArtifactoryArtifact{URL: "https://github.com/argoproj/argo-workflows/not-found"},
 			},
-		}, "/tmp/not-found")
+		}, tempFile)
 		require.Error(t, err)
 		argoError, ok := err.(errors.ArgoError)
 		require.True(t, ok)
 		assert.Equal(t, errors.CodeNotFound, argoError.Code())
 	})
 	t.Run("Found", func(t *testing.T) {
+		tempFile := filepath.Join(tempDir, "found")
 		err := driver.Load(&wfv1.Artifact{
 			ArtifactLocation: wfv1.ArtifactLocation{
 				Artifactory: &wfv1.ArtifactoryArtifact{URL: "https://github.com/argoproj/argo-workflows"},
 			},
-		}, "/tmp/found")
+		}, tempFile)
 		require.NoError(t, err)
-		_, err = os.Stat("/tmp/found")
+		_, err = os.Stat(tempFile)
 		require.NoError(t, err)
 	})
 }
@@ -85,7 +94,7 @@ func TestSaveHTTPArtifactRedirect(t *testing.T) {
 	}
 	defer os.RemoveAll(tempDir) // clean up
 
-	tempFile := path.Join(tempDir, "tmpfile")
+	tempFile := filepath.Join(tempDir, "tmpfile")
 	content := "temporary file's content"
 	if err := os.WriteFile(tempFile, []byte(content), 0o600); err != nil {
 		panic(err)