Bug 1885322: Increase OVN upgrade timeout by 15m

test/e2e/upgrade/upgrade.go

@@ -258,8 +258,18 @@ func clusterUpgrade(f *framework.Framework, c configv1client.Interface, dc dynam
 
 	// this is very long.  We should update the clusteroperator junit to give us a duration.
 	maximumDuration := 150 * time.Minute
-	// if upgrades take longer than this, then we will have a junit marker indicating failure.
-	durationToSoftFailure := 75 * time.Minute
+	baseDurationToSoftFailure := 75 * time.Minute
+	durationToSoftFailure := baseDurationToSoftFailure
+
+	network, err := c.ConfigV1().Networks().Get(context.Background(), "cluster", metav1.GetOptions{})
+	framework.ExpectNoError(err)
+	if network.Status.NetworkType == "OVNKubernetes" {
+		// deploying with OVN is expected to take longer. on average, ~15m longer
+		// some extra context to this increase which links to a jira showing which operators take longer:
+		// compared to OpenShiftSDN:
+		//   https://bugzilla.redhat.com/show_bug.cgi?id=1942164
+		durationToSoftFailure = baseDurationToSoftFailure + (15 * time.Minute)
+	}
 
 	framework.Logf("Starting upgrade to version=%s image=%s", version.Version.String(), version.NodeImage)
 
@@ -271,9 +281,11 @@ func clusterUpgrade(f *framework.Framework, c configv1client.Interface, dc dynam
 	case upgradeAbortAtRandom:
 		abortAt = int(rand.Int31n(100) + 1)
 		maximumDuration *= 2
+		durationToSoftFailure *= 2
 		framework.Logf("Upgrade will be aborted and the cluster will roll back to the current version after %d%% of operators have upgraded (picked randomly)", abortAt)
 	default:
 		maximumDuration *= 2
+		durationToSoftFailure *= 2
 		framework.Logf("Upgrade will be aborted and the cluster will roll back to the current version after %d%% of operators have upgraded", upgradeAbortAt)
 	}
 
@@ -346,6 +358,7 @@ func clusterUpgrade(f *framework.Framework, c configv1client.Interface, dc dynam
 		func() error {
 			framework.Logf("Cluster version operator acknowledged upgrade request")
 			aborted := false
+			action := "upgrade"
 			var lastMessage string
 			upgradeStarted := time.Now()
 
@@ -379,28 +392,30 @@ func clusterUpgrade(f *framework.Framework, c configv1client.Interface, dc dynam
 						return false, err
 					}
 					aborted = true
+					action = "aborted upgrade"
 					return false, nil
 				}
 
 				return monitor.Reached(cv, desired)
 
 			}); err != nil {
 				if lastMessage != "" {
-					return fmt.Errorf("Cluster did not complete upgrade: %v: %s", err, lastMessage)
+					return fmt.Errorf("Cluster did not complete %s: %v: %s", action, err, lastMessage)
 				}
-				return fmt.Errorf("Cluster did not complete upgrade: %v", err)
+				return fmt.Errorf("Cluster did not complete %s: %v", action, err)
 			}
 
-			framework.Logf("Completed upgrade to %s", versionString(desired))
+			framework.Logf("Completed %s to %s", action, versionString(desired))
 
 			// record whether the cluster was fast or slow upgrading.  Don't fail the test, we still want signal on the actual tests themselves.
 			upgradeEnded := time.Now()
 			upgradeDuration := upgradeEnded.Sub(upgradeStarted)
+			testCaseName := fmt.Sprintf("[sig-cluster-lifecycle] cluster upgrade should complete in %0.2f minutes", durationToSoftFailure.Minutes())
+			failure := ""
 			if upgradeDuration > durationToSoftFailure {
-				disruption.RecordJUnitResult(f, "[sig-cluster-lifecycle] cluster upgrade should be fast", upgradeDuration, fmt.Sprintf("Upgrade took too long: %v", upgradeDuration.Minutes()))
-			} else {
-				disruption.RecordJUnitResult(f, "[sig-cluster-lifecycle] cluster upgrade should be fast", upgradeDuration, "")
+				failure = fmt.Sprintf("%s to %s took too long: %0.2f minutes", action, versionString(desired), upgradeDuration.Minutes())
 			}
+			disruption.RecordJUnitResult(f, testCaseName, upgradeDuration, failure)
 
 			return nil
 		},