To record the start and end time of job scheduling

pkg/scheduler/actions/allocate/allocate.go

@@ -17,6 +17,8 @@
 package allocate
 
 import (
+	"time"
+
 	"k8s.io/klog"
 
 	"volcano.sh/volcano/pkg/scheduler/api"
@@ -225,6 +227,7 @@ func (alloc *Action) Execute(ssn *framework.Session) {
 						task.UID, node.Name, ssn.UID, err)
 				} else {
 					metrics.UpdateE2eSchedulingDurationByJob(job.Name, string(job.Queue), job.Namespace, metrics.Duration(job.CreationTimestamp.Time))
+					metrics.UpdateE2eSchedulingLastTimeByJob(job.Name, string(job.Queue), job.Namespace, time.Now())
 				}
 			} else {
 				klog.V(3).Infof("Predicates failed for task <%s/%s> on node <%s> with limited resources",
@@ -239,6 +242,7 @@ func (alloc *Action) Execute(ssn *framework.Session) {
 							task.UID, node.Name, ssn.UID, err)
 					} else {
 						metrics.UpdateE2eSchedulingDurationByJob(job.Name, string(job.Queue), job.Namespace, metrics.Duration(job.CreationTimestamp.Time))
+						metrics.UpdateE2eSchedulingLastTimeByJob(job.Name, string(job.Queue), job.Namespace, time.Now())
 					}
 				}
 			}

pkg/scheduler/actions/backfill/backfill.go

@@ -17,6 +17,8 @@ limitations under the License.
 package backfill
 
 import (
+	"time"
+
 	"k8s.io/klog"
 
 	"volcano.sh/volcano/pkg/scheduler/api"
@@ -76,6 +78,7 @@ func (backfill *Action) Execute(ssn *framework.Session) {
 					}
 
 					metrics.UpdateE2eSchedulingDurationByJob(job.Name, string(job.Queue), job.Namespace, metrics.Duration(job.CreationTimestamp.Time))
+					metrics.UpdateE2eSchedulingLastTimeByJob(job.Name, string(job.Queue), job.Namespace, time.Now())
 					allocated = true
 					break
 				}

pkg/scheduler/cache/event_handlers.go

@@ -420,6 +420,8 @@ func (sc *SchedulerCache) setPodGroup(ss *schedulingapi.PodGroup) error {
 		sc.Jobs[job].Queue = schedulingapi.QueueID(sc.defaultQueue)
 	}
 
+	metrics.UpdateE2eSchedulingStartTimeByJob(sc.Jobs[job].Name, string(sc.Jobs[job].Queue), sc.Jobs[job].Namespace,
+		sc.Jobs[job].CreationTimestamp.Time)
 	return nil
 }
 

pkg/scheduler/metrics/job.go

@@ -52,6 +52,8 @@ func RegisterJobRetries(jobID string) {
 // DeleteJobMetrics delete all metrics related to the job
 func DeleteJobMetrics(jobName, queue, namespace string) {
 	e2eJobSchedulingDuration.DeleteLabelValues(jobName, queue, namespace)
+	e2eJobSchedulingStartTime.DeleteLabelValues(jobName, queue, namespace)
+	e2eJobSchedulingLastTime.DeleteLabelValues(jobName, queue, namespace)
 	unscheduleTaskCount.DeleteLabelValues(jobName)
 	jobShare.DeleteLabelValues(namespace, jobName)
 	jobRetryCount.DeleteLabelValues(jobName)

pkg/scheduler/metrics/metrics.go

@@ -62,6 +62,24 @@ var (
 		[]string{"job_name", "queue", "job_namespace"},
 	)
 
+	e2eJobSchedulingStartTime = promauto.NewGaugeVec(
+		prometheus.GaugeOpts{
+			Subsystem: VolcanoNamespace,
+			Name:      "e2e_job_scheduling_start_time",
+			Help:      "E2E job scheduling start time",
+		},
+		[]string{"job_name", "queue", "job_namespace"},
+	)
+
+	e2eJobSchedulingLastTime = promauto.NewGaugeVec(
+		prometheus.GaugeOpts{
+			Subsystem: VolcanoNamespace,
+			Name:      "e2e_job_scheduling_last_time",
+			Help:      "E2E job scheduling last time",
+		},
+		[]string{"job_name", "queue", "job_namespace"},
+	)
+
 	pluginSchedulingLatency = promauto.NewHistogramVec(
 		prometheus.HistogramOpts{
 			Subsystem: VolcanoNamespace,
@@ -151,6 +169,16 @@ func UpdateE2eSchedulingDurationByJob(jobName string, queue string, namespace st
 	e2eJobSchedulingLatency.Observe(DurationInMilliseconds(duration))
 }
 
+// UpdateE2eSchedulingStartTimeByJob updates the start time of scheduling
+func UpdateE2eSchedulingStartTimeByJob(jobName string, queue string, namespace string, t time.Time) {
+	e2eJobSchedulingStartTime.WithLabelValues(jobName, queue, namespace).Set(ConvertToUnix(t))
+}
+
+// UpdateE2eSchedulingLastTimeByJob updates the last time of scheduling
+func UpdateE2eSchedulingLastTimeByJob(jobName string, queue string, namespace string, t time.Time) {
+	e2eJobSchedulingLastTime.WithLabelValues(jobName, queue, namespace).Set(ConvertToUnix(t))
+}
+
 // UpdateTaskScheduleDuration updates single task scheduling latency
 func UpdateTaskScheduleDuration(duration time.Duration) {
 	taskSchedulingLatency.Observe(DurationInMilliseconds(duration))
@@ -200,3 +228,8 @@ func DurationInSeconds(duration time.Duration) float64 {
 func Duration(start time.Time) time.Duration {
 	return time.Since(start)
 }
+
+// ConvertToUnix convert the time to Unix time
+func ConvertToUnix(t time.Time) float64 {
+	return float64(t.Unix())
+}