1
0
mirror of https://github.com/prometheus/alertmanager.git synced 2026-02-05 15:45:34 +01:00
Files
alertmanager/provider/mem/mem_test.go
Siavash Safi c90d8707d5 feat(provider): implement per-alert limits (#4819)
* feat(limit): add new limit package with bucket

Add a new limit package with generic bucket implementation.
This can be used for example to limit the number of alerts in memory.

Benchmarks:
```go
goos: darwin
goarch: arm64
pkg: github.com/prometheus/alertmanager/limit
cpu: Apple M3 Pro
BenchmarkBucketUpsert/EmptyBucket-12  	 8816954	       122.4 ns/op	      56 B/op	       2 allocs/op
BenchmarkBucketUpsert/AddToFullBucketWithExpiredItems-12         	 9861010	       123.0 ns/op	      56 B/op	       2 allocs/op
BenchmarkBucketUpsert/AddToFullBucketWithActiveItems-12          	 8343778	       143.6 ns/op	      56 B/op	       2 allocs/op
BenchmarkBucketUpsert/UpdateExistingAlert-12                     	10107787	       118.9 ns/op	      56 B/op	       2 allocs/op
BenchmarkBucketUpsert/MixedWorkload-12                           	 9436174	       126.0 ns/op	      56 B/op	       2 allocs/op
BenchmarkBucketUpsertScaling/BucketSize_10-12                    	10255278	       115.4 ns/op	      56 B/op	       2 allocs/op
BenchmarkBucketUpsertScaling/BucketSize_50-12                    	10166518	       117.1 ns/op	      56 B/op	       2 allocs/op
BenchmarkBucketUpsertScaling/BucketSize_100-12                   	10457394	       115.0 ns/op	      56 B/op	       2 allocs/op
BenchmarkBucketUpsertScaling/BucketSize_500-12                   	 9644079	       115.2 ns/op	      56 B/op	       2 allocs/op
BenchmarkBucketUpsertScaling/BucketSize_1000-12                  	10426184	       116.6 ns/op	      56 B/op	       2 allocs/op
BenchmarkBucketUpsertConcurrent-12                               	 5796210	       216.3 ns/op	     406 B/op	       5 allocs/op
PASS
ok  	github.com/prometheus/alertmanager/limit	15.497s
```

Signed-off-by: Siavash Safi <siavash@cloudflare.com>

* feat(provider): implement per-alert limits

Use the new limit module to add optional per alert-name limits.
The metrics for limited alerts can be enabled using
`alerts-limited-metric` feature flag.

Signed-off-by: Siavash Safi <siavash@cloudflare.com>

---------

Signed-off-by: Siavash Safi <siavash@cloudflare.com>
2026-02-01 13:21:20 +01:00

695 lines
19 KiB
Go

// Copyright The Prometheus Authors
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package mem
import (
"context"
"errors"
"fmt"
"reflect"
"strconv"
"sync"
"sync/atomic"
"testing"
"time"
"github.com/prometheus/client_golang/prometheus"
"github.com/prometheus/common/model"
"github.com/prometheus/common/promslog"
"github.com/stretchr/testify/require"
"github.com/prometheus/alertmanager/store"
"github.com/prometheus/alertmanager/types"
)
var (
t0 = time.Now()
t1 = t0.Add(100 * time.Millisecond)
alert1 = &types.Alert{
Alert: model.Alert{
Labels: model.LabelSet{"bar": "foo"},
Annotations: model.LabelSet{"foo": "bar"},
StartsAt: t0,
EndsAt: t1,
GeneratorURL: "http://example.com/prometheus",
},
UpdatedAt: t0,
Timeout: false,
}
alert2 = &types.Alert{
Alert: model.Alert{
Labels: model.LabelSet{"bar": "foo2"},
Annotations: model.LabelSet{"foo": "bar2"},
StartsAt: t0,
EndsAt: t1,
GeneratorURL: "http://example.com/prometheus",
},
UpdatedAt: t0,
Timeout: false,
}
alert3 = &types.Alert{
Alert: model.Alert{
Labels: model.LabelSet{"bar": "foo3"},
Annotations: model.LabelSet{"foo": "bar3"},
StartsAt: t0,
EndsAt: t1,
GeneratorURL: "http://example.com/prometheus",
},
UpdatedAt: t0,
Timeout: false,
}
)
// TestAlertsSubscribePutStarvation tests starvation of `iterator.Close` and
// `alerts.Put`. Both `Subscribe` and `Put` use the Alerts.mtx lock. `Subscribe`
// needs it to subscribe and more importantly unsubscribe `Alerts.listeners`. `Put`
// uses the lock to add additional alerts and iterate the `Alerts.listeners` map.
// If the channel of a listener is at its limit, `alerts.Lock` is blocked, whereby
// a listener can not unsubscribe as the lock is hold by `alerts.Lock`.
func TestAlertsSubscribePutStarvation(t *testing.T) {
marker := types.NewMarker(prometheus.NewRegistry())
alerts, err := NewAlerts(context.Background(), marker, 30*time.Minute, 0, noopCallback{}, promslog.NewNopLogger(), prometheus.NewRegistry(), nil)
if err != nil {
t.Fatal(err)
}
iterator := alerts.Subscribe("test")
alertsToInsert := []*types.Alert{}
// Exhaust alert channel
for i := range alertChannelLength + 1 {
alertsToInsert = append(alertsToInsert, &types.Alert{
Alert: model.Alert{
// Make sure the fingerprints differ
Labels: model.LabelSet{"iteration": model.LabelValue(strconv.Itoa(i))},
Annotations: model.LabelSet{"foo": "bar"},
StartsAt: t0,
EndsAt: t1,
GeneratorURL: "http://example.com/prometheus",
},
UpdatedAt: t0,
Timeout: false,
})
}
putIsDone := make(chan struct{})
putsErr := make(chan error, 1)
go func() {
if err := alerts.Put(context.Background(), alertsToInsert...); err != nil {
putsErr <- err
return
}
putIsDone <- struct{}{}
}()
// Increase probability that `iterator.Close` is called after `alerts.Put`.
time.Sleep(100 * time.Millisecond)
iterator.Close()
select {
case <-putsErr:
t.Fatal(err)
case <-putIsDone:
// continue
case <-time.After(100 * time.Millisecond):
t.Fatal("expected `alerts.Put` and `iterator.Close` not to starve each other")
}
}
func TestDeadLock(t *testing.T) {
t0 := time.Now()
t1 := t0.Add(5 * time.Second)
marker := types.NewMarker(prometheus.NewRegistry())
// Run gc every 5 milliseconds to increase the possibility of a deadlock with Subscribe()
alerts, err := NewAlerts(context.Background(), marker, 5*time.Millisecond, 0, noopCallback{}, promslog.NewNopLogger(), prometheus.NewRegistry(), nil)
if err != nil {
t.Fatal(err)
}
alertsToInsert := []*types.Alert{}
for i := range 200 + 1 {
alertsToInsert = append(alertsToInsert, &types.Alert{
Alert: model.Alert{
// Make sure the fingerprints differ
Labels: model.LabelSet{"iteration": model.LabelValue(strconv.Itoa(i))},
Annotations: model.LabelSet{"foo": "bar"},
StartsAt: t0,
EndsAt: t1,
GeneratorURL: "http://example.com/prometheus",
},
UpdatedAt: t0,
Timeout: false,
})
}
if err := alerts.Put(context.Background(), alertsToInsert...); err != nil {
t.Fatal("Unable to add alerts")
}
done := make(chan bool)
// call subscribe repeatedly in a goroutine to increase
// the possibility of a deadlock occurring
go func() {
tick := time.NewTicker(10 * time.Millisecond)
defer tick.Stop()
stopAfter := time.After(1 * time.Second)
for {
select {
case <-tick.C:
alerts.Subscribe("test")
case <-stopAfter:
done <- true
break
}
}
}()
select {
case <-done:
// no deadlock
alerts.Close()
case <-time.After(10 * time.Second):
t.Error("Deadlock detected")
}
}
func TestAlertsPut(t *testing.T) {
marker := types.NewMarker(prometheus.NewRegistry())
alerts, err := NewAlerts(context.Background(), marker, 30*time.Minute, 0, noopCallback{}, promslog.NewNopLogger(), prometheus.NewRegistry(), nil)
if err != nil {
t.Fatal(err)
}
insert := []*types.Alert{alert1, alert2, alert3}
if err := alerts.Put(context.Background(), insert...); err != nil {
t.Fatalf("Insert failed: %s", err)
}
for i, a := range insert {
res, err := alerts.Get(a.Fingerprint())
if err != nil {
t.Fatalf("retrieval error: %s", err)
}
require.NoError(t, alertDiff(a, res), "unexpected alert: %d", i)
}
}
func TestAlertsSubscribe(t *testing.T) {
marker := types.NewMarker(prometheus.NewRegistry())
ctx := t.Context()
alerts, err := NewAlerts(ctx, marker, 30*time.Minute, 0, noopCallback{}, promslog.NewNopLogger(), prometheus.NewRegistry(), nil)
if err != nil {
t.Fatal(err)
}
// Add alert1 to validate if pending alerts will be sent.
if err := alerts.Put(ctx, alert1); err != nil {
t.Fatalf("Insert failed: %s", err)
}
expectedAlerts := map[model.Fingerprint]*types.Alert{
alert1.Fingerprint(): alert1,
alert2.Fingerprint(): alert2,
alert3.Fingerprint(): alert3,
}
// Start many consumers and make sure that each receives all the subsequent alerts.
var (
nb = 100
fatalc = make(chan string, nb)
wg sync.WaitGroup
)
wg.Add(nb)
for i := range nb {
go func(i int) {
defer wg.Done()
it := alerts.Subscribe("test")
defer it.Close()
received := make(map[model.Fingerprint]struct{})
for {
select {
case got, ok := <-it.Next():
if !ok {
fatalc <- fmt.Sprintf("Iterator %d closed", i)
return
}
if it.Err() != nil {
fatalc <- fmt.Sprintf("Iterator %d: %v", i, it.Err())
return
}
expected := expectedAlerts[got.Data.Fingerprint()]
if err := alertDiff(got.Data, expected); err != nil {
fatalc <- fmt.Sprintf("Unexpected alert (iterator %d)\n%s", i, err.Error())
return
}
received[got.Data.Fingerprint()] = struct{}{}
if len(received) == len(expectedAlerts) {
return
}
case <-time.After(5 * time.Second):
fatalc <- fmt.Sprintf("Unexpected number of alerts for iterator %d, got: %d, expected: %d", i, len(received), len(expectedAlerts))
return
}
}
}(i)
}
// Add more alerts that should be received by the subscribers.
if err := alerts.Put(ctx, alert2); err != nil {
t.Fatalf("Insert failed: %s", err)
}
if err := alerts.Put(ctx, alert3); err != nil {
t.Fatalf("Insert failed: %s", err)
}
wg.Wait()
close(fatalc)
fatal, ok := <-fatalc
if ok {
t.Fatal(fatal)
}
}
func TestAlertsGetPending(t *testing.T) {
marker := types.NewMarker(prometheus.NewRegistry())
alerts, err := NewAlerts(context.Background(), marker, 30*time.Minute, 0, noopCallback{}, promslog.NewNopLogger(), nil, nil)
if err != nil {
t.Fatal(err)
}
ctx := context.Background()
if err := alerts.Put(ctx, alert1, alert2); err != nil {
t.Fatalf("Insert failed: %s", err)
}
expectedAlerts := map[model.Fingerprint]*types.Alert{
alert1.Fingerprint(): alert1,
alert2.Fingerprint(): alert2,
}
iterator := alerts.GetPending()
for actual := range iterator.Next() {
expected := expectedAlerts[actual.Data.Fingerprint()]
require.NoError(t, alertDiff(actual.Data, expected))
}
if err := alerts.Put(ctx, alert3); err != nil {
t.Fatalf("Insert failed: %s", err)
}
expectedAlerts = map[model.Fingerprint]*types.Alert{
alert1.Fingerprint(): alert1,
alert2.Fingerprint(): alert2,
alert3.Fingerprint(): alert3,
}
iterator = alerts.GetPending()
for actual := range iterator.Next() {
expected := expectedAlerts[actual.Data.Fingerprint()]
require.NoError(t, alertDiff(actual.Data, expected))
}
}
func TestAlertsGC(t *testing.T) {
marker := types.NewMarker(prometheus.NewRegistry())
alerts, err := NewAlerts(context.Background(), marker, 200*time.Millisecond, 0, noopCallback{}, promslog.NewNopLogger(), nil, nil)
if err != nil {
t.Fatal(err)
}
insert := []*types.Alert{alert1, alert2, alert3}
if err := alerts.Put(context.Background(), insert...); err != nil {
t.Fatalf("Insert failed: %s", err)
}
for _, a := range insert {
marker.SetActiveOrSilenced(a.Fingerprint(), 0, nil, nil)
marker.SetInhibited(a.Fingerprint())
if !marker.Active(a.Fingerprint()) {
t.Errorf("error setting status: %v", a)
}
}
time.Sleep(300 * time.Millisecond)
for i, a := range insert {
_, err := alerts.Get(a.Fingerprint())
require.Error(t, err)
require.Equal(t, store.ErrNotFound, err, "alert %d didn't get GC'd: %v", i, err)
s := marker.Status(a.Fingerprint())
if s.State != types.AlertStateUnprocessed {
t.Errorf("marker %d didn't get GC'd: %v", i, s)
}
}
}
func TestAlertsStoreCallback(t *testing.T) {
cb := &limitCountCallback{limit: 3}
marker := types.NewMarker(prometheus.NewRegistry())
alerts, err := NewAlerts(context.Background(), marker, 200*time.Millisecond, 0, cb, promslog.NewNopLogger(), nil, nil)
if err != nil {
t.Fatal(err)
}
ctx := context.Background()
err = alerts.Put(ctx, alert1, alert2, alert3)
if err != nil {
t.Fatal(err)
}
if num := cb.alerts.Load(); num != 3 {
t.Fatalf("unexpected number of alerts in the store, expected %v, got %v", 3, num)
}
alert1Mod := *alert1
alert1Mod.Annotations = model.LabelSet{"foo": "bar", "new": "test"} // Update annotations for alert1
alert4 := &types.Alert{
Alert: model.Alert{
Labels: model.LabelSet{"bar4": "foo4"},
Annotations: model.LabelSet{"foo4": "bar4"},
StartsAt: t0,
EndsAt: t1,
GeneratorURL: "http://example.com/prometheus",
},
UpdatedAt: t0,
Timeout: false,
}
err = alerts.Put(ctx, &alert1Mod, alert4)
// Verify that we failed to put new alert into store (not reported via error, only checked using Load)
if err != nil {
t.Fatalf("unexpected error %v", err)
}
if num := cb.alerts.Load(); num != 3 {
t.Fatalf("unexpected number of alerts in the store, expected %v, got %v", 3, num)
}
// But we still managed to update alert1, since callback doesn't report error when updating existing alert.
a, err := alerts.Get(alert1.Fingerprint())
if err != nil {
t.Fatal(err)
}
require.NoError(t, alertDiff(a, &alert1Mod))
// Now wait until existing alerts are GC-ed, and make sure that callback was called.
time.Sleep(300 * time.Millisecond)
if num := cb.alerts.Load(); num != 0 {
t.Fatalf("unexpected number of alerts in the store, expected %v, got %v", 0, num)
}
err = alerts.Put(ctx, alert4)
if err != nil {
t.Fatal(err)
}
}
func TestAlerts_CountByState(t *testing.T) {
marker := types.NewMarker(prometheus.NewRegistry())
alerts, err := NewAlerts(context.Background(), marker, 200*time.Millisecond, 0, nil, promslog.NewNopLogger(), nil, nil)
require.NoError(t, err)
countTotal := func() int {
active, suppressed, unprocessed := alerts.countByState()
return active + suppressed + unprocessed
}
// First, there shouldn't be any alerts.
require.Equal(t, 0, countTotal())
// When you insert a new alert that will eventually be active, it should be unprocessed first.
now := time.Now()
a1 := &types.Alert{
Alert: model.Alert{
Labels: model.LabelSet{"bar": "foo"},
Annotations: model.LabelSet{"foo": "bar"},
StartsAt: now,
EndsAt: now.Add(400 * time.Millisecond),
GeneratorURL: "http://example.com/prometheus",
},
UpdatedAt: now,
Timeout: false,
}
ctx := context.Background()
alerts.Put(ctx, a1)
_, _, unprocessed := alerts.countByState()
require.Equal(t, 1, unprocessed)
require.Equal(t, 1, countTotal())
require.Eventually(t, func() bool {
// When the alert will eventually expire and is considered resolved - it won't count.
return countTotal() == 0
}, 600*time.Millisecond, 100*time.Millisecond)
now = time.Now()
a2 := &types.Alert{
Alert: model.Alert{
Labels: model.LabelSet{"bar": "foo"},
Annotations: model.LabelSet{"foo": "bar"},
StartsAt: now,
EndsAt: now.Add(400 * time.Millisecond),
GeneratorURL: "http://example.com/prometheus",
},
UpdatedAt: now,
Timeout: false,
}
// When insert an alert, and then silence it. It shows up with the correct filter.
alerts.Put(ctx, a2)
marker.SetActiveOrSilenced(a2.Fingerprint(), 1, []string{"1"}, nil)
_, suppressed, _ := alerts.countByState()
require.Equal(t, 1, suppressed)
require.Equal(t, 1, countTotal())
require.Eventually(t, func() bool {
// When the alert will eventually expire and is considered resolved - it won't count.
return countTotal() == 0
}, 600*time.Millisecond, 100*time.Millisecond)
}
func alertDiff(left, right *types.Alert) error {
if left == nil || right == nil {
return errors.New("should not be nil")
}
comparisons := []struct {
name string
isEqual bool
expected any
got any
}{
{"Labels", reflect.DeepEqual(right.Labels, left.Labels), right.Labels, left.Labels},
{"Annotations", reflect.DeepEqual(right.Annotations, left.Annotations), right.Annotations, left.Annotations},
{"StartsAt", right.StartsAt.Equal(left.StartsAt), right.StartsAt, left.StartsAt},
{"EndsAt", right.EndsAt.Equal(left.EndsAt), right.EndsAt, left.EndsAt},
{"UpdatedAt", right.UpdatedAt.Equal(left.UpdatedAt), right.UpdatedAt, left.UpdatedAt},
{"GeneratorURL", right.GeneratorURL == left.GeneratorURL, right.GeneratorURL, left.GeneratorURL},
{"Timeout", right.Timeout == left.Timeout, right.Timeout, left.Timeout},
}
var errs []error
for _, comp := range comparisons {
if !comp.isEqual {
errs = append(errs, fmt.Errorf("field `%s` mismatch.\n Expected: %v\n Got: %v", comp.name, comp.expected, comp.got))
}
}
return errors.Join(errs...)
}
type limitCountCallback struct {
alerts atomic.Int32
limit int
}
var errTooManyAlerts = fmt.Errorf("too many alerts")
func (l *limitCountCallback) PreStore(_ *types.Alert, existing bool) error {
if existing {
return nil
}
if int(l.alerts.Load())+1 > l.limit {
return errTooManyAlerts
}
return nil
}
func (l *limitCountCallback) PostStore(_ *types.Alert, existing bool) {
if !existing {
l.alerts.Add(1)
}
}
func (l *limitCountCallback) PostDelete(_ *types.Alert) {
l.alerts.Add(-1)
}
func TestAlertsConcurrently(t *testing.T) {
callback := &limitCountCallback{limit: 100}
a, err := NewAlerts(context.Background(), types.NewMarker(prometheus.NewRegistry()), time.Millisecond, 0, callback, promslog.NewNopLogger(), nil, nil)
require.NoError(t, err)
stopc := make(chan struct{})
failc := make(chan struct{})
go func() {
time.Sleep(2 * time.Second)
close(stopc)
}()
expire := 10 * time.Millisecond
wg := sync.WaitGroup{}
for range 100 {
wg.Add(1)
go func() {
defer wg.Done()
j := 0
for {
select {
case <-failc:
return
case <-stopc:
return
default:
}
now := time.Now()
err := a.Put(context.Background(), &types.Alert{
Alert: model.Alert{
Labels: model.LabelSet{"bar": model.LabelValue(strconv.Itoa(j))},
StartsAt: now,
EndsAt: now.Add(expire),
},
UpdatedAt: now,
})
if err != nil && !errors.Is(err, errTooManyAlerts) {
close(failc)
return
}
j++
}
}()
}
wg.Wait()
select {
case <-failc:
t.Fatalf("unexpected error happened")
default:
}
time.Sleep(expire)
require.Eventually(t, func() bool {
// When the alert will eventually expire and is considered resolved - it won't count.
active, _, _ := a.countByState()
return active == 0
}, 2*expire, expire)
require.Equal(t, int32(0), callback.alerts.Load())
}
func TestSubscriberChannelMetrics(t *testing.T) {
marker := types.NewMarker(prometheus.NewRegistry())
reg := prometheus.NewRegistry()
alerts, err := NewAlerts(context.Background(), marker, 30*time.Minute, 0, noopCallback{}, promslog.NewNopLogger(), reg, nil)
require.NoError(t, err)
subscriberName := "test_subscriber"
// Subscribe to alerts
iterator := alerts.Subscribe(subscriberName)
defer iterator.Close()
// Consume alerts in the background
go func() {
for range iterator.Next() {
// Just drain the channel
}
}()
// Helper function to get counter value
getCounterValue := func(name, labelName, labelValue string) float64 {
metrics, err := reg.Gather()
require.NoError(t, err)
for _, mf := range metrics {
if mf.GetName() == name {
for _, m := range mf.GetMetric() {
for _, label := range m.GetLabel() {
if label.GetName() == labelName && label.GetValue() == labelValue {
return m.GetCounter().GetValue()
}
}
}
}
}
return 0
}
// Initially, the counter should be 0
writeCount := getCounterValue("alertmanager_alerts_subscriber_channel_writes_total", "subscriber", subscriberName)
require.Equal(t, 0.0, writeCount, "subscriberChannelWrites should start at 0")
// Put some alerts
now := time.Now()
alertsToSend := []*types.Alert{
{
Alert: model.Alert{
Labels: model.LabelSet{"test": "1"},
Annotations: model.LabelSet{"foo": "bar"},
StartsAt: now,
EndsAt: now.Add(1 * time.Hour),
GeneratorURL: "http://example.com/prometheus",
},
UpdatedAt: now,
Timeout: false,
},
{
Alert: model.Alert{
Labels: model.LabelSet{"test": "2"},
Annotations: model.LabelSet{"foo": "bar"},
StartsAt: now,
EndsAt: now.Add(1 * time.Hour),
GeneratorURL: "http://example.com/prometheus",
},
UpdatedAt: now,
Timeout: false,
},
{
Alert: model.Alert{
Labels: model.LabelSet{"test": "3"},
Annotations: model.LabelSet{"foo": "bar"},
StartsAt: now,
EndsAt: now.Add(1 * time.Hour),
GeneratorURL: "http://example.com/prometheus",
},
UpdatedAt: now,
Timeout: false,
},
}
err = alerts.Put(context.Background(), alertsToSend...)
require.NoError(t, err)
// Verify the counter incremented for each successful write
require.Eventually(t, func() bool {
writeCount := getCounterValue("alertmanager_alerts_subscriber_channel_writes_total", "subscriber", subscriberName)
return writeCount == float64(len(alertsToSend))
}, 1*time.Second, 10*time.Millisecond, "subscriberChannelWrites should equal the number of alerts sent")
}