Merge pull request #31343 from mkowalski/fix/haproxy-all-down-flake-fallback

NO-ISSUE: monitor/haproxy: treat all-apiservers-down test as flake

Author: openshift-merge-bot[bot]

pkg/monitortests/network/onpremhaproxy/monitortest.go

@@ -440,7 +440,10 @@ func evaluateFullAPIOutages(downIntervals monitorapi.Intervals) []*junitapi.JUni
 		SystemOut: strings.Join(failures, "\n"),
 	}
 
-	return []*junitapi.JUnitTestCase{failure}
+	// Return both failure and success so the CI system treats this as a flake
+	// rather than a hard failure. This lets us track flake rates in Sippy
+	// without blocking payload acceptance while we tune the detection logic.
+	return []*junitapi.JUnitTestCase{failure, {Name: testName}}
 }
 
 func (w *operatorLogAnalyzer) WriteContentToStorage(ctx context.Context, storageDir, timeSuffix string, finalIntervals monitorapi.Intervals, finalResourceState monitorapi.ResourcesMap) error {

pkg/monitortests/network/onpremhaproxy/monitortest_test.go

@@ -263,14 +263,18 @@ func TestEvaluateFullAPIOutages(t *testing.T) {
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
 			junits := evaluateFullAPIOutages(tt.intervals)
-			require.Len(t, junits, 1)
 
 			if !tt.expectFailure {
+				require.Len(t, junits, 1)
 				assert.Nil(t, junits[0].FailureOutput, "expected the test to pass")
 				return
 			}
 
-			require.NotNil(t, junits[0].FailureOutput, "expected the test to fail")
+			// Failures return both a failure and a success junit so that the CI system
+			// treats the result as a flake rather than a hard failure.
+			require.Len(t, junits, 2, "expected both a failure and a flake-success junit")
+			require.NotNil(t, junits[0].FailureOutput, "expected the first junit to be a failure")
+			assert.Nil(t, junits[1].FailureOutput, "expected the second junit to be a success (flake marker)")
 			for _, expectedOutput := range tt.expectedOutputs {
 				assert.Contains(t, junits[0].SystemOut, expectedOutput)
 			}