Remove --since-time Flag in IPFIX Collector Log Retrieval

This commit addresses the unnecessary use of the --since-time flag when fetching logs from the IPFIX collector Pod in the e2e test. The decision is based on two considerations: first, we already employ the current time for flow record filtering, obviating the need for this flag; second, potential time differences between the testbed and Kubernetes nodes or the time difference between k8s nodes could result in undesired log filtering. Additionally, a 2-second waiting time has been introduced before sending traffic to mitigate time difference issues."

Besides, for flows requiring correlation, such as inter-node traffic, if the records aren't received from both nodes, the record will remain in the queue for up to 3 cycles of aggregatorActiveFlowRecordTimeout. We've set the timeout to exporterActiveFlowExportTimeout plus aggregatorActiveFlowRecordTimeout multiplied by 3 to ensure stability in the interNode test and facilitate the timely observation of removal messages.

Signed-off-by: Yun-Tang Hsu <hsuy@vmware.com>

test/e2e/flowaggregator_test.go

@@ -935,6 +935,8 @@ func checkRecordsForFlows(t *testing.T, data *TestData, srcIP string, dstIP stri
 		cmdStr += fmt.Sprintf(" -p %d", iperfSvcPort)
 	}
 	timeNow := time.Now()
+	// wait 2 seconds to mitigate the possibility that there is a time difference between the K8s Nodes.
+	time.Sleep(2 * time.Second)
 	stdout, _, err := data.RunCommandFromPod(data.testNamespace, "perftest-a", "iperf", []string{"bash", "-c", cmdStr})
 	require.NoErrorf(t, err, "Error when running iperf3 client: %v", err)
 	bwSlice, srcPort, _ := getBandwidthAndPorts(stdout)
@@ -1383,13 +1385,13 @@ func getUint64FieldFromRecord(t *testing.T, record string, field string) uint64
 func getCollectorOutput(t *testing.T, srcIP, dstIP, srcPort string, isDstService bool, checkAllRecords bool, isIPv6 bool, data *TestData, timeSince time.Time) (string, []string) {
 	var collectorOutput string
 	var recordSlices []string
-	// In the ToExternalFlows test, flow record will arrive 5.5s (exporterActiveFlowExportTimeout+aggregatorActiveFlowRecordTimeout) after executing wget command
-	// We set the timeout to 9s (5.5s plus one more aggregatorActiveFlowRecordTimeout) to make the ToExternalFlows test more stable
-	err := wait.PollImmediate(500*time.Millisecond, exporterActiveFlowExportTimeout+aggregatorActiveFlowRecordTimeout*2, func() (bool, error) {
+	// For flows requiring correlation, such as inter-node traffic, if records aren't received from both nodes, the record will remain in the queue for up to 3 cycles of aggregatorActiveFlowRecordTimeout.
+	// To ensure stability in the interNode test and facilitate the timely observation of removal messages, we've set the timeout to exporterActiveFlowExportTimeout plus aggregatorActiveFlowRecordTimeout multiplied by 3.
+	err := wait.PollImmediate(500*time.Millisecond, exporterActiveFlowExportTimeout+aggregatorActiveFlowRecordTimeout*3, func() (bool, error) {
 		var rc int
 		var err error
 		// `pod-running-timeout` option is added to cover scenarios where ipfix flow-collector has crashed after being deployed
-		rc, collectorOutput, _, err = data.RunCommandOnNode(controlPlaneNodeName(), fmt.Sprintf("kubectl logs --pod-running-timeout=%v ipfix-collector -n %s --since-time %s", aggregatorInactiveFlowRecordTimeout.String(), data.testNamespace, timeSince.Format(time.RFC3339)))
+		rc, collectorOutput, _, err = data.RunCommandOnNode(controlPlaneNodeName(), fmt.Sprintf("kubectl logs --pod-running-timeout=%v ipfix-collector -n %s", aggregatorInactiveFlowRecordTimeout.String(), data.testNamespace))
 		if err != nil || rc != 0 {
 			return false, err
 		}