chore: add CI runner disk space cleanup, fix flaky tests (#157)

Author: GrigoryPervakov

.github/workflows/ci.yaml

@@ -190,6 +190,9 @@ jobs:
       - name: Checkout code
         uses: actions/checkout@v6
 
+      - name: Free disk space
+        uses: jlumbroso/free-disk-space@v1.3.0
+
       - name: Set up Go
         uses: actions/setup-go@v6
         with:
@@ -233,7 +236,7 @@ jobs:
           overwrite: true
 
   e2e-test:
-    needs: [ lint, build_and_test ]
+    needs: [ lint ]
     strategy:
       fail-fast: false
       matrix:
@@ -277,13 +280,13 @@ jobs:
     if: ${{ !cancelled() }}
     steps:
       - name: Download all compat test reports
-        uses: actions/download-artifact@v6
+        uses: actions/download-artifact@v8
         with:
           pattern: compat-e2e-report-*
 
 
       - name: Download all e2e test reports
-        uses: actions/download-artifact@v6
+        uses: actions/download-artifact@v8
         with:
           pattern: e2e-report-*
 

.github/workflows/release-main.yml

@@ -19,6 +19,8 @@ jobs:
         uses: actions/checkout@v6
         with:
           fetch-tags: true
+      - name: Free disk space
+        uses: jlumbroso/free-disk-space@v1.3.0
       - name: Set up Go
         uses: actions/setup-go@v6
         with:

.github/workflows/release.yaml

@@ -22,6 +22,8 @@ jobs:
     steps:
       - name: Checkout code
         uses: actions/checkout@v6
+      - name: Free disk space
+        uses: jlumbroso/free-disk-space@v1.3.0
       - name: Log in to the Container registry
         uses: docker/login-action@v4
         with:
@@ -38,6 +40,8 @@ jobs:
     steps:
       - name: Checkout code
         uses: actions/checkout@v6
+      - name: Free disk space
+        uses: jlumbroso/free-disk-space@v1.3.0
       - name: Log in to the Container registry
         uses: docker/login-action@v4
         with:

internal/controller/clickhouse/commands_test.go

@@ -140,6 +140,7 @@ var _ = Describe("commander", Ordered, Label("integration"), func() {
 		})
 
 		chPort := strconv.FormatInt(PortNative, 10) + "/tcp"
+		chHTTPPort := strconv.FormatInt(PortHTTP, 10) + "/tcp"
 		conns := map[v1.ClickHouseReplicaID]clickhouse.Conn{}
 
 		for i := range testReplicas {
@@ -164,8 +165,8 @@ var _ = Describe("commander", Ordered, Label("integration"), func() {
 							FileMode:          0o644,
 						},
 					},
-					ExposedPorts: []string{chPort},
-					WaitingFor:   wait.ForListeningPort(nat.Port(chPort)),
+					ExposedPorts: []string{chPort, chHTTPPort},
+					WaitingFor:   wait.ForHTTP("/").WithPort(nat.Port(chHTTPPort)).WithStartupTimeout(2 * time.Minute),
 				},
 				Started: true,
 			})

test/deploy/deploy_test.go

@@ -223,10 +223,32 @@ var _ = Describe("OLM deployment", Ordered, Label("olm"), func() {
 			_ = testutil.UninstallCRDs(ctx)
 		})
 
+		DeferCleanup(func(ctx context.Context) {
+			if !CurrentSpecReport().Failed() {
+				return
+			}
+
+			By("dumping OLM state for debugging")
+
+			for _, resource := range []string{"catalogsources", "subscriptions", "installplans", "clusterserviceversions"} {
+				out, _ := testutil.Run(exec.CommandContext(ctx, "kubectl", "get", resource,
+					"-n", namespace, "-o", "wide"))
+				AddReportEntry("=== OLM "+resource, string(out))
+			}
+		})
+
 		By("waiting for catalog server to be ready")
 		runCmd(ctx, "kubectl", "wait", "-n", namespace, "--timeout=120s",
 			"--for=condition=Ready", "pod/test-catalog-server")
 
+		By("waiting for ClusterServiceVersion to succeed")
+		Eventually(func(g Gomega) {
+			out, err := testutil.Run(exec.CommandContext(ctx, "kubectl", "get", "csv",
+				"-n", namespace, "--no-headers", "-o", "custom-columns=PHASE:.status.phase"))
+			g.Expect(err).ToNot(HaveOccurred(), string(out))
+			g.Expect(strings.TrimSpace(string(out))).To(Equal("Succeeded"), "CSV phase: %s", strings.TrimSpace(string(out)))
+		}, "5m", "5s").Should(Succeed())
+
 		By("Waiting controller to be ready")
 		Eventually(func(g Gomega) {
 			out, err := testutil.Run(exec.CommandContext(ctx, "kubectl", "wait", "-n", namespace,