mirror of
https://github.com/vmware-tanzu/velero.git
synced 2026-07-26 01:53:09 +00:00
Address review feedback: use DeleteLabelValues, extract resync method
Replace blanket Reset() with targeted DeleteLabelValues to avoid briefly wiping metrics for schedules that still exist. Track known schedules in a set and only delete stale entries on each resync. Extract the wait.Until closure into resyncBackupMetrics() so tests can call it directly without goroutine timing. Replace hand-rolled collectGaugeCount helper with testutil.CollectAndCount. Signed-off-by: Shubham Pampattiwar <spampatt@redhat.com>
This commit is contained in:
@@ -107,10 +107,11 @@ type backupReconciler struct {
|
||||
credentialFileStore credentials.FileStore
|
||||
maxConcurrentK8SConnections int
|
||||
defaultSnapshotMoveData bool
|
||||
globalCRClient kbclient.Client
|
||||
itemBlockWorkerCount int
|
||||
concurrentBackups int
|
||||
globalVolumePoliciesConfigMap string
|
||||
globalCRClient kbclient.Client
|
||||
itemBlockWorkerCount int
|
||||
concurrentBackups int
|
||||
globalVolumePoliciesConfigMap string
|
||||
knownSchedulesWithSuccessfulBackup sets.Set[string]
|
||||
}
|
||||
|
||||
func NewBackupReconciler(
|
||||
@@ -204,30 +205,43 @@ func (b *backupReconciler) updateTotalBackupMetric() {
|
||||
time.Sleep(5 * time.Second)
|
||||
|
||||
wait.Until(
|
||||
func() {
|
||||
// recompute backup_total metric
|
||||
backups := &velerov1api.BackupList{}
|
||||
err := b.kbClient.List(context.Background(), backups, &kbclient.ListOptions{LabelSelector: labels.Everything()})
|
||||
if err != nil {
|
||||
b.logger.Error(err, "Error computing backup_total metric")
|
||||
} else {
|
||||
b.metrics.SetBackupTotal(int64(len(backups.Items)))
|
||||
|
||||
// recompute backup_last_successful_timestamp metric for each
|
||||
// schedule (including the empty schedule, i.e. ad-hoc backups).
|
||||
// Reset first to prune stale entries for deleted schedules.
|
||||
b.metrics.ResetBackupLastSuccessfulTimestamp()
|
||||
for schedule, timestamp := range getLastSuccessBySchedule(backups.Items) {
|
||||
b.metrics.SetBackupLastSuccessfulTimestamp(schedule, timestamp)
|
||||
}
|
||||
}
|
||||
},
|
||||
b.resyncBackupMetrics,
|
||||
backupResyncPeriod,
|
||||
b.ctx.Done(),
|
||||
)
|
||||
}()
|
||||
}
|
||||
|
||||
func (b *backupReconciler) resyncBackupMetrics() {
|
||||
backups := &velerov1api.BackupList{}
|
||||
err := b.kbClient.List(context.Background(), backups, &kbclient.ListOptions{LabelSelector: labels.Everything()})
|
||||
if err != nil {
|
||||
b.logger.Error(err, "Error computing backup_total metric")
|
||||
return
|
||||
}
|
||||
|
||||
b.metrics.SetBackupTotal(int64(len(backups.Items)))
|
||||
|
||||
currentSchedules := getLastSuccessBySchedule(backups.Items)
|
||||
for schedule, timestamp := range currentSchedules {
|
||||
b.metrics.SetBackupLastSuccessfulTimestamp(schedule, timestamp)
|
||||
}
|
||||
|
||||
// Remove metrics for schedules that no longer have successful backups
|
||||
if b.knownSchedulesWithSuccessfulBackup != nil {
|
||||
for schedule := range b.knownSchedulesWithSuccessfulBackup {
|
||||
if _, exists := currentSchedules[schedule]; !exists {
|
||||
b.metrics.DeleteBackupLastSuccessfulTimestamp(schedule)
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
b.knownSchedulesWithSuccessfulBackup = sets.New[string]()
|
||||
for schedule := range currentSchedules {
|
||||
b.knownSchedulesWithSuccessfulBackup.Insert(schedule)
|
||||
}
|
||||
}
|
||||
|
||||
// getLastSuccessBySchedule finds the most recent completed backup for each schedule
|
||||
// and returns a map of schedule name -> completion time of the most recent completed
|
||||
// backup. This map includes an entry for ad-hoc/non-scheduled backups, where the key
|
||||
|
||||
@@ -18,7 +18,6 @@ package controller
|
||||
|
||||
import (
|
||||
"bytes"
|
||||
"context"
|
||||
"fmt"
|
||||
"io"
|
||||
"reflect"
|
||||
@@ -32,7 +31,7 @@ import (
|
||||
"github.com/google/go-cmp/cmp"
|
||||
"github.com/google/go-cmp/cmp/cmpopts"
|
||||
snapshotv1api "github.com/kubernetes-csi/external-snapshotter/client/v8/apis/volumesnapshot/v1"
|
||||
"github.com/prometheus/client_golang/prometheus"
|
||||
"github.com/prometheus/client_golang/prometheus/testutil"
|
||||
"github.com/sirupsen/logrus"
|
||||
"github.com/stretchr/testify/assert"
|
||||
"github.com/stretchr/testify/mock"
|
||||
@@ -2043,60 +2042,15 @@ func Test_getLastSuccessBySchedule(t *testing.T) {
|
||||
}
|
||||
}
|
||||
|
||||
// Test_updateTotalBackupMetric_prunesStaleTimestamps verifies that the periodic
|
||||
// resync removes backupLastSuccessfulTimestamp entries for schedules that no longer
|
||||
// have any completed backups (e.g. after the schedule and its backups are deleted).
|
||||
func Test_updateTotalBackupMetric_prunesStaleTimestamps(t *testing.T) {
|
||||
// Test_resyncBackupMetrics_prunesStaleTimestamps verifies that resyncBackupMetrics
|
||||
// removes backupLastSuccessfulTimestamp entries for schedules that no longer have
|
||||
// any completed backups (e.g. after the schedule and its backups are deleted).
|
||||
func Test_resyncBackupMetrics_prunesStaleTimestamps(t *testing.T) {
|
||||
baseTime, err := time.Parse(time.RFC1123, time.RFC1123)
|
||||
require.NoError(t, err)
|
||||
|
||||
m := metrics.NewServerMetrics()
|
||||
gauge := m.Metrics()["backup_last_successful_timestamp"].(*prometheus.GaugeVec)
|
||||
|
||||
// Simulate a previous resync that set the metric for "deleted-schedule"
|
||||
m.SetBackupLastSuccessfulTimestamp("deleted-schedule", baseTime)
|
||||
require.Equal(t, 1, collectGaugeCount(t, gauge))
|
||||
|
||||
// Current backups only contain entries for "active-schedule"
|
||||
backups := []velerov1api.Backup{
|
||||
*builder.ForBackup("velero", "b1").
|
||||
ObjectMeta(builder.WithLabels(velerov1api.ScheduleNameLabel, "active-schedule")).
|
||||
Phase(velerov1api.BackupPhaseCompleted).
|
||||
CompletionTimestamp(baseTime).
|
||||
Result(),
|
||||
}
|
||||
|
||||
// Replicate the resync logic: reset then set
|
||||
m.ResetBackupLastSuccessfulTimestamp()
|
||||
for schedule, timestamp := range getLastSuccessBySchedule(backups) {
|
||||
m.SetBackupLastSuccessfulTimestamp(schedule, timestamp)
|
||||
}
|
||||
|
||||
// Only "active-schedule" should remain; "deleted-schedule" should be pruned
|
||||
assert.Equal(t, 1, collectGaugeCount(t, gauge))
|
||||
}
|
||||
|
||||
func collectGaugeCount(t *testing.T, g *prometheus.GaugeVec) int {
|
||||
t.Helper()
|
||||
ch := make(chan prometheus.Metric, 10)
|
||||
g.Collect(ch)
|
||||
close(ch)
|
||||
count := 0
|
||||
for range ch {
|
||||
count++
|
||||
}
|
||||
return count
|
||||
}
|
||||
|
||||
// Test_updateTotalBackupMetric_prunesStaleTimestamps_integration tests the actual
|
||||
// updateTotalBackupMetric goroutine with a fake client to verify stale metrics are
|
||||
// pruned during a real resync cycle.
|
||||
func Test_updateTotalBackupMetric_prunesStaleTimestamps_integration(t *testing.T) {
|
||||
baseTime, err := time.Parse(time.RFC1123, time.RFC1123)
|
||||
require.NoError(t, err)
|
||||
|
||||
m := metrics.NewServerMetrics()
|
||||
gauge := m.Metrics()["backup_last_successful_timestamp"].(*prometheus.GaugeVec)
|
||||
gauge := m.Metrics()["backup_last_successful_timestamp"]
|
||||
|
||||
activeBackup := builder.ForBackup("velero", "b1").
|
||||
ObjectMeta(builder.WithLabels(velerov1api.ScheduleNameLabel, "active-schedule")).
|
||||
@@ -2104,26 +2058,30 @@ func Test_updateTotalBackupMetric_prunesStaleTimestamps_integration(t *testing.T
|
||||
CompletionTimestamp(baseTime).
|
||||
Result()
|
||||
|
||||
fakeClient := velerotest.NewFakeControllerRuntimeClient(t, activeBackup)
|
||||
deletedBackup := builder.ForBackup("velero", "b2").
|
||||
ObjectMeta(builder.WithLabels(velerov1api.ScheduleNameLabel, "deleted-schedule")).
|
||||
Phase(velerov1api.BackupPhaseCompleted).
|
||||
CompletionTimestamp(baseTime).
|
||||
Result()
|
||||
|
||||
m.SetBackupLastSuccessfulTimestamp("deleted-schedule", baseTime)
|
||||
require.Equal(t, 1, collectGaugeCount(t, gauge))
|
||||
|
||||
ctx, cancel := context.WithCancel(context.Background())
|
||||
defer cancel()
|
||||
fakeClient := velerotest.NewFakeControllerRuntimeClient(t, activeBackup, deletedBackup)
|
||||
|
||||
c := &backupReconciler{
|
||||
ctx: ctx,
|
||||
kbClient: fakeClient,
|
||||
logger: logrus.StandardLogger(),
|
||||
metrics: m,
|
||||
}
|
||||
|
||||
c.updateTotalBackupMetric()
|
||||
time.Sleep(7 * time.Second)
|
||||
cancel()
|
||||
// First resync: sets metrics for both schedules
|
||||
c.resyncBackupMetrics()
|
||||
assert.Equal(t, 2, testutil.CollectAndCount(gauge))
|
||||
|
||||
assert.Equal(t, 1, collectGaugeCount(t, gauge))
|
||||
// Simulate schedule deletion: remove the backup for "deleted-schedule"
|
||||
require.NoError(t, fakeClient.Delete(t.Context(), deletedBackup))
|
||||
|
||||
// Second resync: prunes "deleted-schedule" metric, keeps "active-schedule"
|
||||
c.resyncBackupMetrics()
|
||||
assert.Equal(t, 1, testutil.CollectAndCount(gauge))
|
||||
}
|
||||
|
||||
// Unit tests to make sure that the backup's status is updated correctly during reconcile.
|
||||
|
||||
@@ -758,10 +758,11 @@ func (m *ServerMetrics) RegisterPodVolumeOpLatencyGauge(node, pvbName, opName, b
|
||||
}
|
||||
}
|
||||
|
||||
// ResetBackupLastSuccessfulTimestamp removes all schedule-level backupLastSuccessfulTimestamp values.
|
||||
func (m *ServerMetrics) ResetBackupLastSuccessfulTimestamp() {
|
||||
// DeleteBackupLastSuccessfulTimestamp removes the backupLastSuccessfulTimestamp
|
||||
// metric for a single schedule.
|
||||
func (m *ServerMetrics) DeleteBackupLastSuccessfulTimestamp(scheduleName string) {
|
||||
if g, ok := m.metrics[backupLastSuccessfulTimestamp].(*prometheus.GaugeVec); ok {
|
||||
g.Reset()
|
||||
g.DeleteLabelValues(scheduleName)
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@@ -21,6 +21,7 @@ import (
|
||||
"time"
|
||||
|
||||
"github.com/prometheus/client_golang/prometheus"
|
||||
"github.com/prometheus/client_golang/prometheus/testutil"
|
||||
dto "github.com/prometheus/client_model/go"
|
||||
"github.com/stretchr/testify/assert"
|
||||
"github.com/stretchr/testify/require"
|
||||
@@ -457,9 +458,9 @@ func getHistogramCount(t *testing.T, vec *prometheus.HistogramVec, scheduleLabel
|
||||
return 0
|
||||
}
|
||||
|
||||
// TestResetBackupLastSuccessfulTimestamp verifies that ResetBackupLastSuccessfulTimestamp
|
||||
// removes all schedule-level values from the backupLastSuccessfulTimestamp gauge.
|
||||
func TestResetBackupLastSuccessfulTimestamp(t *testing.T) {
|
||||
// TestDeleteBackupLastSuccessfulTimestamp verifies that DeleteBackupLastSuccessfulTimestamp
|
||||
// removes only the specified schedule's metric.
|
||||
func TestDeleteBackupLastSuccessfulTimestamp(t *testing.T) {
|
||||
m := NewServerMetrics()
|
||||
|
||||
now := time.Now()
|
||||
@@ -467,13 +468,17 @@ func TestResetBackupLastSuccessfulTimestamp(t *testing.T) {
|
||||
m.SetBackupLastSuccessfulTimestamp("schedule-2", now.Add(-time.Hour))
|
||||
m.SetBackupLastSuccessfulTimestamp("", now.Add(-2*time.Hour))
|
||||
|
||||
// Verify all three entries exist
|
||||
g := m.metrics[backupLastSuccessfulTimestamp].(*prometheus.GaugeVec)
|
||||
assert.Equal(t, 3, collectGaugeCount(t, g))
|
||||
assert.Equal(t, 3, testutil.CollectAndCount(g))
|
||||
|
||||
// Reset should remove all entries
|
||||
m.ResetBackupLastSuccessfulTimestamp()
|
||||
assert.Equal(t, 0, collectGaugeCount(t, g))
|
||||
m.DeleteBackupLastSuccessfulTimestamp("schedule-1")
|
||||
assert.Equal(t, 2, testutil.CollectAndCount(g))
|
||||
|
||||
m.DeleteBackupLastSuccessfulTimestamp("schedule-2")
|
||||
assert.Equal(t, 1, testutil.CollectAndCount(g))
|
||||
|
||||
m.DeleteBackupLastSuccessfulTimestamp("")
|
||||
assert.Equal(t, 0, testutil.CollectAndCount(g))
|
||||
}
|
||||
|
||||
// collectGaugeCount returns the number of time series in a GaugeVec.
|
||||
|
||||
Reference in New Issue
Block a user