VictoriaMetrics/app/vmalert/rule/alerting_test.go

1154 lines
38 KiB
Go
Raw Normal View History

package rule
import (
"context"
"errors"
"fmt"
"reflect"
"sort"
"strings"
"sync"
"testing"
"time"
"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/config"
"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/datasource"
"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/notifier"
"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/utils"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutils"
)
func TestAlertingRule_ToTimeSeries(t *testing.T) {
timestamp := time.Now()
testCases := []struct {
rule *AlertingRule
alert *notifier.Alert
expTS []prompbmarshal.TimeSeries
}{
{
newTestAlertingRule("instant", 0),
&notifier.Alert{State: notifier.StateFiring},
[]prompbmarshal.TimeSeries{
newTimeSeries([]float64{1}, []int64{timestamp.UnixNano()}, map[string]string{
"__name__": alertMetricName,
alertStateLabel: notifier.StateFiring.String(),
}),
},
},
{
newTestAlertingRule("instant extra labels", 0),
&notifier.Alert{State: notifier.StateFiring, Labels: map[string]string{
"job": "foo",
"instance": "bar",
}},
[]prompbmarshal.TimeSeries{
newTimeSeries([]float64{1}, []int64{timestamp.UnixNano()}, map[string]string{
"__name__": alertMetricName,
alertStateLabel: notifier.StateFiring.String(),
"job": "foo",
"instance": "bar",
}),
},
},
{
newTestAlertingRule("instant labels override", 0),
&notifier.Alert{State: notifier.StateFiring, Labels: map[string]string{
alertStateLabel: "foo",
"__name__": "bar",
}},
[]prompbmarshal.TimeSeries{
newTimeSeries([]float64{1}, []int64{timestamp.UnixNano()}, map[string]string{
"__name__": alertMetricName,
alertStateLabel: notifier.StateFiring.String(),
}),
},
},
{
newTestAlertingRule("for", time.Second),
Vmalert compliance 2 (#2340) * vmalert: split alert's `Start` field into `ActiveAt` and `Start` The `ActiveAt` field identifies when alert becomes active for rules with `for > 0`. Previously, this value was stored in field `Start`. The field `Start` now identifies the moment alert became `FIRING`. The split is needed in order to distinguish these two moments in the API responses for alerts. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: support specific moment of time for rules evaluation The Querier interface was extended to accept a new argument used as a timestamp at which evaluation should be made. It is needed to align rules execution time within the group. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: mark disappeared series as stale Series generated by alerting rules, which were sent to remote write now will be marked as stale if they will disappear on the next evaluation. This would make ALERTS and ALERTS_FOR_TIME series more precise. Signed-off-by: hagen1778 <roman@victoriametrics.com> * wip Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: evaluate rules at fixed timestamp Before, time at which rules were evaluated was calculated right before rule execution. The change makes sure that timestamp is calculated only once per evalution round and all rules are using the same timestamp. It also updates the logic of resending of already resolved alert notification. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: allow overridin `alertname` label value if it is present in response Previously, `alertname` was always equal to the Alerting Rule name. Now, its value can be overriden if series in response containt the different value for this label. The change is needed for improving compatibility with Prometheus. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: align rules evaluation in time Now, evaluation timestamp for rules evaluates as if there was no delay in rules evaluation. It means, that rules will be evaluated at fixed timestamps+group_interval. This way provides more consistent evaluation results and improves compatibility with Prometheus, Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: add metric for missed iterations New metric `vmalert_iteration_missed_total` will show whether rules evaluation round was missed. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: reduce delay before the initial rule evaluation in group Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: rollback alertname override According to the spec: ``` The alert name from the alerting rule (HighRequestLatency from the example above) MUST be added to the labels of the alert with the label name as alertname. It MUST override any existing alertname label. ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-3 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: throw err immediately on dedup detection ``` The execution of an alerting rule MUST error out immediately and MUST NOT send any alerts or add samples to samples receiver if there is more than one alert with the same labels ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-4 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: cleanup Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: use strings builder to reduce allocs Signed-off-by: hagen1778 <roman@victoriametrics.com>
2022-03-29 15:09:07 +02:00
&notifier.Alert{State: notifier.StateFiring, ActiveAt: timestamp.Add(time.Second)},
[]prompbmarshal.TimeSeries{
newTimeSeries([]float64{1}, []int64{timestamp.UnixNano()}, map[string]string{
"__name__": alertMetricName,
alertStateLabel: notifier.StateFiring.String(),
}),
newTimeSeries([]float64{float64(timestamp.Add(time.Second).Unix())},
[]int64{timestamp.UnixNano()},
map[string]string{
"__name__": alertForStateMetricName,
}),
},
},
{
newTestAlertingRule("for pending", 10*time.Second),
Vmalert compliance 2 (#2340) * vmalert: split alert's `Start` field into `ActiveAt` and `Start` The `ActiveAt` field identifies when alert becomes active for rules with `for > 0`. Previously, this value was stored in field `Start`. The field `Start` now identifies the moment alert became `FIRING`. The split is needed in order to distinguish these two moments in the API responses for alerts. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: support specific moment of time for rules evaluation The Querier interface was extended to accept a new argument used as a timestamp at which evaluation should be made. It is needed to align rules execution time within the group. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: mark disappeared series as stale Series generated by alerting rules, which were sent to remote write now will be marked as stale if they will disappear on the next evaluation. This would make ALERTS and ALERTS_FOR_TIME series more precise. Signed-off-by: hagen1778 <roman@victoriametrics.com> * wip Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: evaluate rules at fixed timestamp Before, time at which rules were evaluated was calculated right before rule execution. The change makes sure that timestamp is calculated only once per evalution round and all rules are using the same timestamp. It also updates the logic of resending of already resolved alert notification. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: allow overridin `alertname` label value if it is present in response Previously, `alertname` was always equal to the Alerting Rule name. Now, its value can be overriden if series in response containt the different value for this label. The change is needed for improving compatibility with Prometheus. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: align rules evaluation in time Now, evaluation timestamp for rules evaluates as if there was no delay in rules evaluation. It means, that rules will be evaluated at fixed timestamps+group_interval. This way provides more consistent evaluation results and improves compatibility with Prometheus, Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: add metric for missed iterations New metric `vmalert_iteration_missed_total` will show whether rules evaluation round was missed. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: reduce delay before the initial rule evaluation in group Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: rollback alertname override According to the spec: ``` The alert name from the alerting rule (HighRequestLatency from the example above) MUST be added to the labels of the alert with the label name as alertname. It MUST override any existing alertname label. ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-3 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: throw err immediately on dedup detection ``` The execution of an alerting rule MUST error out immediately and MUST NOT send any alerts or add samples to samples receiver if there is more than one alert with the same labels ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-4 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: cleanup Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: use strings builder to reduce allocs Signed-off-by: hagen1778 <roman@victoriametrics.com>
2022-03-29 15:09:07 +02:00
&notifier.Alert{State: notifier.StatePending, ActiveAt: timestamp.Add(time.Second)},
[]prompbmarshal.TimeSeries{
newTimeSeries([]float64{1}, []int64{timestamp.UnixNano()}, map[string]string{
"__name__": alertMetricName,
alertStateLabel: notifier.StatePending.String(),
}),
newTimeSeries([]float64{float64(timestamp.Add(time.Second).Unix())},
[]int64{timestamp.UnixNano()},
map[string]string{
"__name__": alertForStateMetricName,
}),
},
},
}
for _, tc := range testCases {
t.Run(tc.rule.Name, func(t *testing.T) {
tc.rule.alerts[tc.alert.ID] = tc.alert
tss := tc.rule.toTimeSeries(timestamp.Unix())
if err := compareTimeSeries(t, tc.expTS, tss); err != nil {
t.Fatalf("timeseries missmatch: %s", err)
}
})
}
}
func TestAlertingRule_Exec(t *testing.T) {
const defaultStep = 5 * time.Millisecond
type testAlert struct {
labels []string
alert *notifier.Alert
}
testCases := []struct {
rule *AlertingRule
steps [][]datasource.Metric
expAlerts map[int][]testAlert
}{
{
newTestAlertingRule("empty", 0),
[][]datasource.Metric{},
nil,
},
{
newTestAlertingRule("empty labels", 0),
[][]datasource.Metric{
{datasource.Metric{Values: []float64{1}, Timestamps: []int64{1}}},
},
map[int][]testAlert{
0: {{alert: &notifier.Alert{State: notifier.StateFiring}}},
},
},
{
Vmalert compliance 2 (#2340) * vmalert: split alert's `Start` field into `ActiveAt` and `Start` The `ActiveAt` field identifies when alert becomes active for rules with `for > 0`. Previously, this value was stored in field `Start`. The field `Start` now identifies the moment alert became `FIRING`. The split is needed in order to distinguish these two moments in the API responses for alerts. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: support specific moment of time for rules evaluation The Querier interface was extended to accept a new argument used as a timestamp at which evaluation should be made. It is needed to align rules execution time within the group. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: mark disappeared series as stale Series generated by alerting rules, which were sent to remote write now will be marked as stale if they will disappear on the next evaluation. This would make ALERTS and ALERTS_FOR_TIME series more precise. Signed-off-by: hagen1778 <roman@victoriametrics.com> * wip Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: evaluate rules at fixed timestamp Before, time at which rules were evaluated was calculated right before rule execution. The change makes sure that timestamp is calculated only once per evalution round and all rules are using the same timestamp. It also updates the logic of resending of already resolved alert notification. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: allow overridin `alertname` label value if it is present in response Previously, `alertname` was always equal to the Alerting Rule name. Now, its value can be overriden if series in response containt the different value for this label. The change is needed for improving compatibility with Prometheus. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: align rules evaluation in time Now, evaluation timestamp for rules evaluates as if there was no delay in rules evaluation. It means, that rules will be evaluated at fixed timestamps+group_interval. This way provides more consistent evaluation results and improves compatibility with Prometheus, Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: add metric for missed iterations New metric `vmalert_iteration_missed_total` will show whether rules evaluation round was missed. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: reduce delay before the initial rule evaluation in group Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: rollback alertname override According to the spec: ``` The alert name from the alerting rule (HighRequestLatency from the example above) MUST be added to the labels of the alert with the label name as alertname. It MUST override any existing alertname label. ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-3 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: throw err immediately on dedup detection ``` The execution of an alerting rule MUST error out immediately and MUST NOT send any alerts or add samples to samples receiver if there is more than one alert with the same labels ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-4 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: cleanup Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: use strings builder to reduce allocs Signed-off-by: hagen1778 <roman@victoriametrics.com>
2022-03-29 15:09:07 +02:00
newTestAlertingRule("single-firing=>inactive=>firing=>inactive=>inactive", 0),
[][]datasource.Metric{
{metricWithLabels(t, "name", "foo")},
{},
{metricWithLabels(t, "name", "foo")},
{},
{},
},
map[int][]testAlert{
0: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateFiring}}},
1: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateInactive}}},
2: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateFiring}}},
3: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateInactive}}},
4: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateInactive}}},
Vmalert compliance 2 (#2340) * vmalert: split alert's `Start` field into `ActiveAt` and `Start` The `ActiveAt` field identifies when alert becomes active for rules with `for > 0`. Previously, this value was stored in field `Start`. The field `Start` now identifies the moment alert became `FIRING`. The split is needed in order to distinguish these two moments in the API responses for alerts. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: support specific moment of time for rules evaluation The Querier interface was extended to accept a new argument used as a timestamp at which evaluation should be made. It is needed to align rules execution time within the group. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: mark disappeared series as stale Series generated by alerting rules, which were sent to remote write now will be marked as stale if they will disappear on the next evaluation. This would make ALERTS and ALERTS_FOR_TIME series more precise. Signed-off-by: hagen1778 <roman@victoriametrics.com> * wip Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: evaluate rules at fixed timestamp Before, time at which rules were evaluated was calculated right before rule execution. The change makes sure that timestamp is calculated only once per evalution round and all rules are using the same timestamp. It also updates the logic of resending of already resolved alert notification. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: allow overridin `alertname` label value if it is present in response Previously, `alertname` was always equal to the Alerting Rule name. Now, its value can be overriden if series in response containt the different value for this label. The change is needed for improving compatibility with Prometheus. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: align rules evaluation in time Now, evaluation timestamp for rules evaluates as if there was no delay in rules evaluation. It means, that rules will be evaluated at fixed timestamps+group_interval. This way provides more consistent evaluation results and improves compatibility with Prometheus, Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: add metric for missed iterations New metric `vmalert_iteration_missed_total` will show whether rules evaluation round was missed. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: reduce delay before the initial rule evaluation in group Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: rollback alertname override According to the spec: ``` The alert name from the alerting rule (HighRequestLatency from the example above) MUST be added to the labels of the alert with the label name as alertname. It MUST override any existing alertname label. ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-3 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: throw err immediately on dedup detection ``` The execution of an alerting rule MUST error out immediately and MUST NOT send any alerts or add samples to samples receiver if there is more than one alert with the same labels ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-4 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: cleanup Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: use strings builder to reduce allocs Signed-off-by: hagen1778 <roman@victoriametrics.com>
2022-03-29 15:09:07 +02:00
},
},
{
newTestAlertingRule("single-firing=>inactive=>firing=>inactive=>inactive=>firing", 0),
[][]datasource.Metric{
{metricWithLabels(t, "name", "foo")},
{},
{metricWithLabels(t, "name", "foo")},
{},
{},
{metricWithLabels(t, "name", "foo")},
},
map[int][]testAlert{
0: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateFiring}}},
1: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateInactive}}},
2: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateFiring}}},
3: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateInactive}}},
4: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateInactive}}},
5: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateFiring}}},
},
},
{
newTestAlertingRule("multiple-firing", 0),
[][]datasource.Metric{
{
metricWithLabels(t, "name", "foo"),
metricWithLabels(t, "name", "foo1"),
metricWithLabels(t, "name", "foo2"),
},
},
map[int][]testAlert{
0: {
{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateFiring}},
{labels: []string{"name", "foo1"}, alert: &notifier.Alert{State: notifier.StateFiring}},
{labels: []string{"name", "foo2"}, alert: &notifier.Alert{State: notifier.StateFiring}},
},
},
},
{
newTestAlertingRule("multiple-steps-firing", 0),
[][]datasource.Metric{
{metricWithLabels(t, "name", "foo")},
{metricWithLabels(t, "name", "foo1")},
{metricWithLabels(t, "name", "foo2")},
},
// 1: fire first alert
// 2: fire second alert, set first inactive
Vmalert compliance 2 (#2340) * vmalert: split alert's `Start` field into `ActiveAt` and `Start` The `ActiveAt` field identifies when alert becomes active for rules with `for > 0`. Previously, this value was stored in field `Start`. The field `Start` now identifies the moment alert became `FIRING`. The split is needed in order to distinguish these two moments in the API responses for alerts. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: support specific moment of time for rules evaluation The Querier interface was extended to accept a new argument used as a timestamp at which evaluation should be made. It is needed to align rules execution time within the group. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: mark disappeared series as stale Series generated by alerting rules, which were sent to remote write now will be marked as stale if they will disappear on the next evaluation. This would make ALERTS and ALERTS_FOR_TIME series more precise. Signed-off-by: hagen1778 <roman@victoriametrics.com> * wip Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: evaluate rules at fixed timestamp Before, time at which rules were evaluated was calculated right before rule execution. The change makes sure that timestamp is calculated only once per evalution round and all rules are using the same timestamp. It also updates the logic of resending of already resolved alert notification. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: allow overridin `alertname` label value if it is present in response Previously, `alertname` was always equal to the Alerting Rule name. Now, its value can be overriden if series in response containt the different value for this label. The change is needed for improving compatibility with Prometheus. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: align rules evaluation in time Now, evaluation timestamp for rules evaluates as if there was no delay in rules evaluation. It means, that rules will be evaluated at fixed timestamps+group_interval. This way provides more consistent evaluation results and improves compatibility with Prometheus, Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: add metric for missed iterations New metric `vmalert_iteration_missed_total` will show whether rules evaluation round was missed. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: reduce delay before the initial rule evaluation in group Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: rollback alertname override According to the spec: ``` The alert name from the alerting rule (HighRequestLatency from the example above) MUST be added to the labels of the alert with the label name as alertname. It MUST override any existing alertname label. ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-3 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: throw err immediately on dedup detection ``` The execution of an alerting rule MUST error out immediately and MUST NOT send any alerts or add samples to samples receiver if there is more than one alert with the same labels ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-4 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: cleanup Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: use strings builder to reduce allocs Signed-off-by: hagen1778 <roman@victoriametrics.com>
2022-03-29 15:09:07 +02:00
// 3: fire third alert, set second inactive
map[int][]testAlert{
0: {
{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateFiring}},
},
1: {
{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateInactive}},
{labels: []string{"name", "foo1"}, alert: &notifier.Alert{State: notifier.StateFiring}},
},
2: {
{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateInactive}},
{labels: []string{"name", "foo1"}, alert: &notifier.Alert{State: notifier.StateInactive}},
{labels: []string{"name", "foo2"}, alert: &notifier.Alert{State: notifier.StateFiring}},
},
},
},
{
newTestAlertingRule("for-pending", time.Minute),
[][]datasource.Metric{
{metricWithLabels(t, "name", "foo")},
},
map[int][]testAlert{
0: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StatePending}}},
},
},
{
newTestAlertingRule("for-fired", defaultStep),
[][]datasource.Metric{
{metricWithLabels(t, "name", "foo")},
{metricWithLabels(t, "name", "foo")},
},
map[int][]testAlert{
0: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StatePending}}},
1: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateFiring}}},
},
},
{
newTestAlertingRule("for-pending=>empty", time.Second),
[][]datasource.Metric{
{metricWithLabels(t, "name", "foo")},
{metricWithLabels(t, "name", "foo")},
// empty step to delete pending alerts
{},
},
map[int][]testAlert{
0: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StatePending}}},
1: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StatePending}}},
2: {},
},
},
{
newTestAlertingRule("for-pending=>firing=>inactive=>pending=>firing", defaultStep),
[][]datasource.Metric{
{metricWithLabels(t, "name", "foo")},
{metricWithLabels(t, "name", "foo")},
// empty step to set alert inactive
{},
{metricWithLabels(t, "name", "foo")},
{metricWithLabels(t, "name", "foo")},
},
map[int][]testAlert{
0: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StatePending}}},
1: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateFiring}}},
2: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateInactive}}},
3: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StatePending}}},
4: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateFiring}}},
},
},
{
newTestAlertingRuleWithKeepFiring("for-pending=>firing=>keepfiring=>firing", defaultStep, defaultStep),
[][]datasource.Metric{
{metricWithLabels(t, "name", "foo")},
{metricWithLabels(t, "name", "foo")},
// empty step to keep firing
{},
{metricWithLabels(t, "name", "foo")},
},
map[int][]testAlert{
0: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StatePending}}},
1: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateFiring}}},
2: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateFiring}}},
3: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateFiring}}},
},
},
{
newTestAlertingRuleWithKeepFiring("for-pending=>firing=>keepfiring=>keepfiring=>inactive=>pending=>firing", defaultStep, 2*defaultStep),
[][]datasource.Metric{
{metricWithLabels(t, "name", "foo")},
{metricWithLabels(t, "name", "foo")},
// empty step to keep firing
{},
// another empty step to keep firing
{},
// empty step to set alert inactive
{},
{metricWithLabels(t, "name", "foo")},
{metricWithLabels(t, "name", "foo")},
},
map[int][]testAlert{
0: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StatePending}}},
1: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateFiring}}},
2: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateFiring}}},
3: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateFiring}}},
4: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateInactive}}},
5: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StatePending}}},
6: {{labels: []string{"name", "foo"}, alert: &notifier.Alert{State: notifier.StateFiring}}},
},
},
}
fakeGroup := Group{Name: "TestRule_Exec"}
for _, tc := range testCases {
t.Run(tc.rule.Name, func(t *testing.T) {
fq := &datasource.FakeQuerier{}
tc.rule.q = fq
tc.rule.GroupID = fakeGroup.ID()
for i, step := range tc.steps {
fq.Reset()
fq.Add(step...)
if _, err := tc.rule.exec(context.TODO(), time.Now(), 0); err != nil {
t.Fatalf("unexpected err: %s", err)
}
// artificial delay between applying steps
time.Sleep(defaultStep)
if _, ok := tc.expAlerts[i]; !ok {
continue
}
if len(tc.rule.alerts) != len(tc.expAlerts[i]) {
t.Fatalf("evalIndex %d: expected %d alerts; got %d", i, len(tc.expAlerts[i]), len(tc.rule.alerts))
}
expAlerts := make(map[uint64]*notifier.Alert)
for _, ta := range tc.expAlerts[i] {
labels := make(map[string]string)
for i := 0; i < len(ta.labels); i += 2 {
k, v := ta.labels[i], ta.labels[i+1]
labels[k] = v
}
labels[alertNameLabel] = tc.rule.Name
h := hash(labels)
expAlerts[h] = ta.alert
}
for key, exp := range expAlerts {
got, ok := tc.rule.alerts[key]
if !ok {
t.Fatalf("evalIndex %d: expected to have key %d", i, key)
}
if got.State != exp.State {
t.Fatalf("evalIndex %d: expected state %d; got %d", i, exp.State, got.State)
}
}
}
})
}
}
func TestAlertingRule_ExecRange(t *testing.T) {
fakeGroup := Group{Name: "TestRule_ExecRange"}
testCases := []struct {
rule *AlertingRule
data []datasource.Metric
expAlerts []*notifier.Alert
expHoldAlertStateAlerts map[uint64]*notifier.Alert
}{
{
newTestAlertingRule("empty", 0),
[]datasource.Metric{},
nil,
nil,
},
{
newTestAlertingRule("empty labels", 0),
[]datasource.Metric{
{Values: []float64{1}, Timestamps: []int64{1}},
},
[]*notifier.Alert{
{State: notifier.StateFiring},
},
nil,
},
{
newTestAlertingRule("single-firing", 0),
[]datasource.Metric{
metricWithLabels(t, "name", "foo"),
},
[]*notifier.Alert{
{
Labels: map[string]string{"name": "foo"},
State: notifier.StateFiring,
},
},
nil,
},
{
newTestAlertingRule("single-firing-on-range", 0),
[]datasource.Metric{
{Values: []float64{1, 1, 1}, Timestamps: []int64{1e3, 2e3, 3e3}},
},
[]*notifier.Alert{
{State: notifier.StateFiring},
{State: notifier.StateFiring},
{State: notifier.StateFiring},
},
nil,
},
{
newTestAlertingRule("for-pending", time.Second),
[]datasource.Metric{
{Values: []float64{1, 1, 1}, Timestamps: []int64{1, 3, 5}},
},
[]*notifier.Alert{
Vmalert compliance 2 (#2340) * vmalert: split alert's `Start` field into `ActiveAt` and `Start` The `ActiveAt` field identifies when alert becomes active for rules with `for > 0`. Previously, this value was stored in field `Start`. The field `Start` now identifies the moment alert became `FIRING`. The split is needed in order to distinguish these two moments in the API responses for alerts. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: support specific moment of time for rules evaluation The Querier interface was extended to accept a new argument used as a timestamp at which evaluation should be made. It is needed to align rules execution time within the group. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: mark disappeared series as stale Series generated by alerting rules, which were sent to remote write now will be marked as stale if they will disappear on the next evaluation. This would make ALERTS and ALERTS_FOR_TIME series more precise. Signed-off-by: hagen1778 <roman@victoriametrics.com> * wip Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: evaluate rules at fixed timestamp Before, time at which rules were evaluated was calculated right before rule execution. The change makes sure that timestamp is calculated only once per evalution round and all rules are using the same timestamp. It also updates the logic of resending of already resolved alert notification. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: allow overridin `alertname` label value if it is present in response Previously, `alertname` was always equal to the Alerting Rule name. Now, its value can be overriden if series in response containt the different value for this label. The change is needed for improving compatibility with Prometheus. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: align rules evaluation in time Now, evaluation timestamp for rules evaluates as if there was no delay in rules evaluation. It means, that rules will be evaluated at fixed timestamps+group_interval. This way provides more consistent evaluation results and improves compatibility with Prometheus, Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: add metric for missed iterations New metric `vmalert_iteration_missed_total` will show whether rules evaluation round was missed. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: reduce delay before the initial rule evaluation in group Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: rollback alertname override According to the spec: ``` The alert name from the alerting rule (HighRequestLatency from the example above) MUST be added to the labels of the alert with the label name as alertname. It MUST override any existing alertname label. ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-3 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: throw err immediately on dedup detection ``` The execution of an alerting rule MUST error out immediately and MUST NOT send any alerts or add samples to samples receiver if there is more than one alert with the same labels ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-4 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: cleanup Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: use strings builder to reduce allocs Signed-off-by: hagen1778 <roman@victoriametrics.com>
2022-03-29 15:09:07 +02:00
{State: notifier.StatePending, ActiveAt: time.Unix(1, 0)},
{State: notifier.StatePending, ActiveAt: time.Unix(3, 0)},
{State: notifier.StatePending, ActiveAt: time.Unix(5, 0)},
},
map[uint64]*notifier.Alert{hash(map[string]string{"alertname": "for-pending"}): {
GroupID: fakeGroup.ID(),
Name: "for-pending",
Labels: map[string]string{"alertname": "for-pending"},
Annotations: map[string]string{},
State: notifier.StatePending,
ActiveAt: time.Unix(5, 0),
Value: 1,
For: time.Second,
}},
},
{
newTestAlertingRule("for-firing", 3*time.Second),
[]datasource.Metric{
{Values: []float64{1, 1, 1}, Timestamps: []int64{1, 3, 5}},
},
[]*notifier.Alert{
Vmalert compliance 2 (#2340) * vmalert: split alert's `Start` field into `ActiveAt` and `Start` The `ActiveAt` field identifies when alert becomes active for rules with `for > 0`. Previously, this value was stored in field `Start`. The field `Start` now identifies the moment alert became `FIRING`. The split is needed in order to distinguish these two moments in the API responses for alerts. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: support specific moment of time for rules evaluation The Querier interface was extended to accept a new argument used as a timestamp at which evaluation should be made. It is needed to align rules execution time within the group. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: mark disappeared series as stale Series generated by alerting rules, which were sent to remote write now will be marked as stale if they will disappear on the next evaluation. This would make ALERTS and ALERTS_FOR_TIME series more precise. Signed-off-by: hagen1778 <roman@victoriametrics.com> * wip Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: evaluate rules at fixed timestamp Before, time at which rules were evaluated was calculated right before rule execution. The change makes sure that timestamp is calculated only once per evalution round and all rules are using the same timestamp. It also updates the logic of resending of already resolved alert notification. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: allow overridin `alertname` label value if it is present in response Previously, `alertname` was always equal to the Alerting Rule name. Now, its value can be overriden if series in response containt the different value for this label. The change is needed for improving compatibility with Prometheus. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: align rules evaluation in time Now, evaluation timestamp for rules evaluates as if there was no delay in rules evaluation. It means, that rules will be evaluated at fixed timestamps+group_interval. This way provides more consistent evaluation results and improves compatibility with Prometheus, Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: add metric for missed iterations New metric `vmalert_iteration_missed_total` will show whether rules evaluation round was missed. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: reduce delay before the initial rule evaluation in group Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: rollback alertname override According to the spec: ``` The alert name from the alerting rule (HighRequestLatency from the example above) MUST be added to the labels of the alert with the label name as alertname. It MUST override any existing alertname label. ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-3 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: throw err immediately on dedup detection ``` The execution of an alerting rule MUST error out immediately and MUST NOT send any alerts or add samples to samples receiver if there is more than one alert with the same labels ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-4 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: cleanup Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: use strings builder to reduce allocs Signed-off-by: hagen1778 <roman@victoriametrics.com>
2022-03-29 15:09:07 +02:00
{State: notifier.StatePending, ActiveAt: time.Unix(1, 0)},
{State: notifier.StatePending, ActiveAt: time.Unix(1, 0)},
{State: notifier.StateFiring, ActiveAt: time.Unix(1, 0)},
},
map[uint64]*notifier.Alert{hash(map[string]string{"alertname": "for-firing"}): {
GroupID: fakeGroup.ID(),
Name: "for-firing",
Labels: map[string]string{"alertname": "for-firing"},
Annotations: map[string]string{},
State: notifier.StateFiring,
ActiveAt: time.Unix(1, 0),
Start: time.Unix(5, 0),
Value: 1,
For: 3 * time.Second,
}},
},
{
newTestAlertingRule("for-hold-pending", time.Second),
[]datasource.Metric{
{Values: []float64{1, 1, 1}, Timestamps: []int64{1, 2, 5}},
},
[]*notifier.Alert{
{State: notifier.StatePending, ActiveAt: time.Unix(1, 0)},
{State: notifier.StateFiring, ActiveAt: time.Unix(1, 0)},
{State: notifier.StatePending, ActiveAt: time.Unix(5, 0)},
},
map[uint64]*notifier.Alert{hash(map[string]string{"alertname": "for-hold-pending"}): {
GroupID: fakeGroup.ID(),
Name: "for-hold-pending",
Labels: map[string]string{"alertname": "for-hold-pending"},
Annotations: map[string]string{},
State: notifier.StatePending,
ActiveAt: time.Unix(5, 0),
Value: 1,
For: time.Second,
}},
},
{
newTestAlertingRule("for=>pending=>firing=>pending=>firing=>pending", time.Second),
[]datasource.Metric{
{Values: []float64{1, 1, 1, 1, 1}, Timestamps: []int64{1, 2, 5, 6, 20}},
},
[]*notifier.Alert{
Vmalert compliance 2 (#2340) * vmalert: split alert's `Start` field into `ActiveAt` and `Start` The `ActiveAt` field identifies when alert becomes active for rules with `for > 0`. Previously, this value was stored in field `Start`. The field `Start` now identifies the moment alert became `FIRING`. The split is needed in order to distinguish these two moments in the API responses for alerts. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: support specific moment of time for rules evaluation The Querier interface was extended to accept a new argument used as a timestamp at which evaluation should be made. It is needed to align rules execution time within the group. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: mark disappeared series as stale Series generated by alerting rules, which were sent to remote write now will be marked as stale if they will disappear on the next evaluation. This would make ALERTS and ALERTS_FOR_TIME series more precise. Signed-off-by: hagen1778 <roman@victoriametrics.com> * wip Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: evaluate rules at fixed timestamp Before, time at which rules were evaluated was calculated right before rule execution. The change makes sure that timestamp is calculated only once per evalution round and all rules are using the same timestamp. It also updates the logic of resending of already resolved alert notification. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: allow overridin `alertname` label value if it is present in response Previously, `alertname` was always equal to the Alerting Rule name. Now, its value can be overriden if series in response containt the different value for this label. The change is needed for improving compatibility with Prometheus. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: align rules evaluation in time Now, evaluation timestamp for rules evaluates as if there was no delay in rules evaluation. It means, that rules will be evaluated at fixed timestamps+group_interval. This way provides more consistent evaluation results and improves compatibility with Prometheus, Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: add metric for missed iterations New metric `vmalert_iteration_missed_total` will show whether rules evaluation round was missed. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: reduce delay before the initial rule evaluation in group Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: rollback alertname override According to the spec: ``` The alert name from the alerting rule (HighRequestLatency from the example above) MUST be added to the labels of the alert with the label name as alertname. It MUST override any existing alertname label. ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-3 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: throw err immediately on dedup detection ``` The execution of an alerting rule MUST error out immediately and MUST NOT send any alerts or add samples to samples receiver if there is more than one alert with the same labels ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-4 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: cleanup Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: use strings builder to reduce allocs Signed-off-by: hagen1778 <roman@victoriametrics.com>
2022-03-29 15:09:07 +02:00
{State: notifier.StatePending, ActiveAt: time.Unix(1, 0)},
{State: notifier.StateFiring, ActiveAt: time.Unix(1, 0)},
{State: notifier.StatePending, ActiveAt: time.Unix(5, 0)},
{State: notifier.StateFiring, ActiveAt: time.Unix(5, 0)},
{State: notifier.StatePending, ActiveAt: time.Unix(20, 0)},
},
nil,
},
{
newTestAlertingRule("multi-series", 3*time.Second),
[]datasource.Metric{
{Values: []float64{1, 1, 1}, Timestamps: []int64{1, 3, 5}},
{
Values: []float64{1, 1}, Timestamps: []int64{1, 5},
Labels: []datasource.Label{{Name: "foo", Value: "bar"}},
},
},
[]*notifier.Alert{
Vmalert compliance 2 (#2340) * vmalert: split alert's `Start` field into `ActiveAt` and `Start` The `ActiveAt` field identifies when alert becomes active for rules with `for > 0`. Previously, this value was stored in field `Start`. The field `Start` now identifies the moment alert became `FIRING`. The split is needed in order to distinguish these two moments in the API responses for alerts. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: support specific moment of time for rules evaluation The Querier interface was extended to accept a new argument used as a timestamp at which evaluation should be made. It is needed to align rules execution time within the group. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: mark disappeared series as stale Series generated by alerting rules, which were sent to remote write now will be marked as stale if they will disappear on the next evaluation. This would make ALERTS and ALERTS_FOR_TIME series more precise. Signed-off-by: hagen1778 <roman@victoriametrics.com> * wip Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: evaluate rules at fixed timestamp Before, time at which rules were evaluated was calculated right before rule execution. The change makes sure that timestamp is calculated only once per evalution round and all rules are using the same timestamp. It also updates the logic of resending of already resolved alert notification. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: allow overridin `alertname` label value if it is present in response Previously, `alertname` was always equal to the Alerting Rule name. Now, its value can be overriden if series in response containt the different value for this label. The change is needed for improving compatibility with Prometheus. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: align rules evaluation in time Now, evaluation timestamp for rules evaluates as if there was no delay in rules evaluation. It means, that rules will be evaluated at fixed timestamps+group_interval. This way provides more consistent evaluation results and improves compatibility with Prometheus, Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: add metric for missed iterations New metric `vmalert_iteration_missed_total` will show whether rules evaluation round was missed. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: reduce delay before the initial rule evaluation in group Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: rollback alertname override According to the spec: ``` The alert name from the alerting rule (HighRequestLatency from the example above) MUST be added to the labels of the alert with the label name as alertname. It MUST override any existing alertname label. ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-3 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: throw err immediately on dedup detection ``` The execution of an alerting rule MUST error out immediately and MUST NOT send any alerts or add samples to samples receiver if there is more than one alert with the same labels ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-4 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: cleanup Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: use strings builder to reduce allocs Signed-off-by: hagen1778 <roman@victoriametrics.com>
2022-03-29 15:09:07 +02:00
{State: notifier.StatePending, ActiveAt: time.Unix(1, 0)},
{State: notifier.StatePending, ActiveAt: time.Unix(1, 0)},
{State: notifier.StateFiring, ActiveAt: time.Unix(1, 0)},
{
State: notifier.StatePending, ActiveAt: time.Unix(1, 0),
Labels: map[string]string{
"foo": "bar",
},
},
{
State: notifier.StatePending, ActiveAt: time.Unix(5, 0),
Labels: map[string]string{
"foo": "bar",
},
},
},
map[uint64]*notifier.Alert{
hash(map[string]string{"alertname": "multi-series"}): {
GroupID: fakeGroup.ID(),
Name: "multi-series",
Labels: map[string]string{"alertname": "multi-series"},
Annotations: map[string]string{},
State: notifier.StateFiring,
ActiveAt: time.Unix(1, 0),
Start: time.Unix(5, 0),
Value: 1,
For: 3 * time.Second,
},
hash(map[string]string{"alertname": "multi-series", "foo": "bar"}): {
GroupID: fakeGroup.ID(),
Name: "multi-series",
Labels: map[string]string{"alertname": "multi-series", "foo": "bar"},
Annotations: map[string]string{},
State: notifier.StatePending,
ActiveAt: time.Unix(5, 0),
Value: 1,
For: 3 * time.Second,
},
},
},
{
newTestRuleWithLabels("multi-series-firing", "source", "vm"),
[]datasource.Metric{
{Values: []float64{1, 1}, Timestamps: []int64{1, 100}},
{
Values: []float64{1, 1}, Timestamps: []int64{1, 5},
Labels: []datasource.Label{{Name: "foo", Value: "bar"}},
},
},
[]*notifier.Alert{
{State: notifier.StateFiring, Labels: map[string]string{
"source": "vm",
}},
{State: notifier.StateFiring, Labels: map[string]string{
"source": "vm",
}},
//
{State: notifier.StateFiring, Labels: map[string]string{
"foo": "bar",
"source": "vm",
}},
{State: notifier.StateFiring, Labels: map[string]string{
"foo": "bar",
"source": "vm",
}},
},
nil,
},
}
for _, tc := range testCases {
t.Run(tc.rule.Name, func(t *testing.T) {
fq := &datasource.FakeQuerier{}
tc.rule.q = fq
tc.rule.GroupID = fakeGroup.ID()
fq.Add(tc.data...)
gotTS, err := tc.rule.execRange(context.TODO(), time.Unix(1, 0), time.Unix(5, 0))
if err != nil {
t.Fatalf("unexpected err: %s", err)
}
var expTS []prompbmarshal.TimeSeries
var j int
for _, series := range tc.data {
for _, timestamp := range series.Timestamps {
a := tc.expAlerts[j]
if a.Labels == nil {
a.Labels = make(map[string]string)
}
a.Labels[alertNameLabel] = tc.rule.Name
Vmalert compliance 2 (#2340) * vmalert: split alert's `Start` field into `ActiveAt` and `Start` The `ActiveAt` field identifies when alert becomes active for rules with `for > 0`. Previously, this value was stored in field `Start`. The field `Start` now identifies the moment alert became `FIRING`. The split is needed in order to distinguish these two moments in the API responses for alerts. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: support specific moment of time for rules evaluation The Querier interface was extended to accept a new argument used as a timestamp at which evaluation should be made. It is needed to align rules execution time within the group. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: mark disappeared series as stale Series generated by alerting rules, which were sent to remote write now will be marked as stale if they will disappear on the next evaluation. This would make ALERTS and ALERTS_FOR_TIME series more precise. Signed-off-by: hagen1778 <roman@victoriametrics.com> * wip Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: evaluate rules at fixed timestamp Before, time at which rules were evaluated was calculated right before rule execution. The change makes sure that timestamp is calculated only once per evalution round and all rules are using the same timestamp. It also updates the logic of resending of already resolved alert notification. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: allow overridin `alertname` label value if it is present in response Previously, `alertname` was always equal to the Alerting Rule name. Now, its value can be overriden if series in response containt the different value for this label. The change is needed for improving compatibility with Prometheus. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: align rules evaluation in time Now, evaluation timestamp for rules evaluates as if there was no delay in rules evaluation. It means, that rules will be evaluated at fixed timestamps+group_interval. This way provides more consistent evaluation results and improves compatibility with Prometheus, Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: add metric for missed iterations New metric `vmalert_iteration_missed_total` will show whether rules evaluation round was missed. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: reduce delay before the initial rule evaluation in group Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: rollback alertname override According to the spec: ``` The alert name from the alerting rule (HighRequestLatency from the example above) MUST be added to the labels of the alert with the label name as alertname. It MUST override any existing alertname label. ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-3 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: throw err immediately on dedup detection ``` The execution of an alerting rule MUST error out immediately and MUST NOT send any alerts or add samples to samples receiver if there is more than one alert with the same labels ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-4 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: cleanup Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: use strings builder to reduce allocs Signed-off-by: hagen1778 <roman@victoriametrics.com>
2022-03-29 15:09:07 +02:00
expTS = append(expTS, tc.rule.alertToTimeSeries(a, timestamp)...)
j++
}
}
if len(gotTS) != len(expTS) {
t.Fatalf("expected %d time series; got %d", len(expTS), len(gotTS))
}
for i := range expTS {
got, exp := gotTS[i], expTS[i]
if !reflect.DeepEqual(got, exp) {
t.Fatalf("%d: expected \n%v but got \n%v", i, exp, got)
}
}
if tc.expHoldAlertStateAlerts != nil {
if !reflect.DeepEqual(tc.expHoldAlertStateAlerts, tc.rule.alerts) {
t.Fatalf("expected hold alerts state: \n%v but got \n%v", tc.expHoldAlertStateAlerts, tc.rule.alerts)
}
}
})
}
}
func TestGroup_Restore(t *testing.T) {
defaultTS := time.Now()
fqr := &datasource.FakeQuerierWithRegistry{}
fn := func(rules []config.Rule, expAlerts map[uint64]*notifier.Alert) {
t.Helper()
defer fqr.Reset()
for _, r := range rules {
fqr.Set(r.Expr, metricWithValueAndLabels(t, 0, "__name__", r.Alert))
}
fg := NewGroup(config.Group{Name: "TestRestore", Rules: rules}, fqr, time.Second, nil)
wg := sync.WaitGroup{}
wg.Add(1)
go func() {
nts := func() []notifier.Notifier { return []notifier.Notifier{&notifier.FakeNotifier{}} }
fg.Start(context.Background(), nts, nil, fqr)
wg.Done()
}()
fg.Close()
wg.Wait()
gotAlerts := make(map[uint64]*notifier.Alert)
for _, rs := range fg.Rules {
alerts := rs.(*AlertingRule).alerts
for k, v := range alerts {
if !v.Restored {
// set not restored alerts to predictable timestamp
v.ActiveAt = defaultTS
}
gotAlerts[k] = v
}
}
if len(gotAlerts) != len(expAlerts) {
t.Fatalf("expected %d alerts; got %d", len(expAlerts), len(gotAlerts))
}
for key, exp := range expAlerts {
got, ok := gotAlerts[key]
if !ok {
t.Fatalf("expected to have key %d", key)
}
if got.State != notifier.StatePending {
t.Fatalf("expected state %d; got %d", notifier.StatePending, got.State)
}
if got.ActiveAt != exp.ActiveAt {
t.Fatalf("expected ActiveAt %v; got %v", exp.ActiveAt, got.ActiveAt)
}
if got.Name != exp.Name {
t.Fatalf("expected alertname %q; got %q", exp.Name, got.Name)
}
}
}
stateMetric := func(name string, value time.Time, labels ...string) datasource.Metric {
labels = append(labels, "__name__", alertForStateMetricName)
labels = append(labels, alertNameLabel, name)
labels = append(labels, alertGroupNameLabel, "TestRestore")
return metricWithValueAndLabels(t, float64(value.Unix()), labels...)
}
// one active alert, no previous state
fn(
[]config.Rule{{Alert: "foo", Expr: "foo", For: promutils.NewDuration(time.Second)}},
map[uint64]*notifier.Alert{
hash(map[string]string{alertNameLabel: "foo", alertGroupNameLabel: "TestRestore"}): {
Name: "foo",
ActiveAt: defaultTS,
},
})
fqr.Reset()
// one active alert with state restore
ts := time.Now().Truncate(time.Hour)
fqr.Set(`last_over_time(ALERTS_FOR_STATE{alertgroup="TestRestore",alertname="foo"}[3600s])`,
stateMetric("foo", ts))
fn(
[]config.Rule{{Alert: "foo", Expr: "foo", For: promutils.NewDuration(time.Second)}},
map[uint64]*notifier.Alert{
hash(map[string]string{alertNameLabel: "foo", alertGroupNameLabel: "TestRestore"}): {
Name: "foo",
ActiveAt: ts,
},
})
// two rules, two active alerts, one with state restored
ts = time.Now().Truncate(time.Hour)
fqr.Set(`last_over_time(ALERTS_FOR_STATE{alertgroup="TestRestore",alertname="bar"}[3600s])`,
stateMetric("bar", ts))
fn(
[]config.Rule{
{Alert: "foo", Expr: "foo", For: promutils.NewDuration(time.Second)},
{Alert: "bar", Expr: "bar", For: promutils.NewDuration(time.Second)},
},
map[uint64]*notifier.Alert{
hash(map[string]string{alertNameLabel: "foo", alertGroupNameLabel: "TestRestore"}): {
Name: "foo",
ActiveAt: defaultTS,
},
hash(map[string]string{alertNameLabel: "bar", alertGroupNameLabel: "TestRestore"}): {
Name: "bar",
ActiveAt: ts,
},
})
// two rules, two active alerts, two with state restored
ts = time.Now().Truncate(time.Hour)
fqr.Set(`last_over_time(ALERTS_FOR_STATE{alertgroup="TestRestore",alertname="foo"}[3600s])`,
stateMetric("foo", ts))
fqr.Set(`last_over_time(ALERTS_FOR_STATE{alertgroup="TestRestore",alertname="bar"}[3600s])`,
stateMetric("bar", ts))
fn(
[]config.Rule{
{Alert: "foo", Expr: "foo", For: promutils.NewDuration(time.Second)},
{Alert: "bar", Expr: "bar", For: promutils.NewDuration(time.Second)},
},
map[uint64]*notifier.Alert{
hash(map[string]string{alertNameLabel: "foo", alertGroupNameLabel: "TestRestore"}): {
Name: "foo",
ActiveAt: ts,
},
hash(map[string]string{alertNameLabel: "bar", alertGroupNameLabel: "TestRestore"}): {
Name: "bar",
ActiveAt: ts,
},
})
// one active alert but wrong state restore
ts = time.Now().Truncate(time.Hour)
fqr.Set(`last_over_time(ALERTS_FOR_STATE{alertname="bar",alertgroup="TestRestore"}[3600s])`,
stateMetric("wrong alert", ts))
fn(
[]config.Rule{{Alert: "foo", Expr: "foo", For: promutils.NewDuration(time.Second)}},
map[uint64]*notifier.Alert{
hash(map[string]string{alertNameLabel: "foo", alertGroupNameLabel: "TestRestore"}): {
Name: "foo",
ActiveAt: defaultTS,
},
})
// one active alert with labels
ts = time.Now().Truncate(time.Hour)
fqr.Set(`last_over_time(ALERTS_FOR_STATE{alertgroup="TestRestore",alertname="foo",env="dev"}[3600s])`,
stateMetric("foo", ts, "env", "dev"))
fn(
[]config.Rule{{Alert: "foo", Expr: "foo", Labels: map[string]string{"env": "dev"}, For: promutils.NewDuration(time.Second)}},
map[uint64]*notifier.Alert{
hash(map[string]string{alertNameLabel: "foo", alertGroupNameLabel: "TestRestore", "env": "dev"}): {
Name: "foo",
ActiveAt: ts,
},
})
// one active alert with restore labels missmatch
ts = time.Now().Truncate(time.Hour)
fqr.Set(`last_over_time(ALERTS_FOR_STATE{alertgroup="TestRestore",alertname="foo",env="dev"}[3600s])`,
stateMetric("foo", ts, "env", "dev", "team", "foo"))
fn(
[]config.Rule{{Alert: "foo", Expr: "foo", Labels: map[string]string{"env": "dev"}, For: promutils.NewDuration(time.Second)}},
map[uint64]*notifier.Alert{
hash(map[string]string{alertNameLabel: "foo", alertGroupNameLabel: "TestRestore", "env": "dev"}): {
Name: "foo",
ActiveAt: defaultTS,
},
})
}
func TestAlertingRule_Exec_Negative(t *testing.T) {
fq := &datasource.FakeQuerier{}
ar := newTestAlertingRule("test", 0)
ar.Labels = map[string]string{"job": "test"}
ar.q = fq
// successful attempt
// label `job` will be overridden by rule extra label, the original value will be reserved by "exported_job"
fq.Add(metricWithValueAndLabels(t, 1, "__name__", "foo", "job", "bar"))
fq.Add(metricWithValueAndLabels(t, 1, "__name__", "foo", "job", "baz"))
_, err := ar.exec(context.TODO(), time.Now(), 0)
if err != nil {
t.Fatal(err)
}
// label `__name__` will be omitted and get duplicated results here
fq.Add(metricWithValueAndLabels(t, 1, "__name__", "foo_1", "job", "bar"))
_, err = ar.exec(context.TODO(), time.Now(), 0)
if !errors.Is(err, errDuplicate) {
t.Fatalf("expected to have %s error; got %s", errDuplicate, err)
}
fq.Reset()
expErr := "connection reset by peer"
fq.SetErr(errors.New(expErr))
_, err = ar.exec(context.TODO(), time.Now(), 0)
if err == nil {
t.Fatalf("expected to get err; got nil")
}
if !strings.Contains(err.Error(), expErr) {
t.Fatalf("expected to get err %q; got %q insterad", expErr, err)
}
}
func TestAlertingRuleLimit(t *testing.T) {
fq := &datasource.FakeQuerier{}
ar := newTestAlertingRule("test", 0)
ar.Labels = map[string]string{"job": "test"}
ar.q = fq
ar.For = time.Minute
testCases := []struct {
limit int
err string
tssNum int
}{
{
limit: 0,
tssNum: 4,
},
{
limit: -1,
tssNum: 4,
},
{
limit: 1,
err: "exec exceeded limit of 1 with 2 alerts",
tssNum: 0,
},
{
limit: 4,
tssNum: 4,
},
}
var (
err error
timestamp = time.Now()
)
fq.Add(metricWithValueAndLabels(t, 1, "__name__", "foo", "job", "bar"))
fq.Add(metricWithValueAndLabels(t, 1, "__name__", "foo", "bar", "job"))
for _, testCase := range testCases {
_, err = ar.exec(context.TODO(), timestamp, testCase.limit)
if err != nil && !strings.EqualFold(err.Error(), testCase.err) {
t.Fatal(err)
}
}
fq.Reset()
}
func TestAlertingRule_Template(t *testing.T) {
testCases := []struct {
rule *AlertingRule
metrics []datasource.Metric
expAlerts map[uint64]*notifier.Alert
}{
{
&AlertingRule{
Name: "common",
Labels: map[string]string{
"region": "east",
},
Annotations: map[string]string{
"summary": `{{ $labels.alertname }}: Too high connection number for "{{ $labels.instance }}"`,
},
alerts: make(map[uint64]*notifier.Alert),
},
[]datasource.Metric{
metricWithValueAndLabels(t, 1, "instance", "foo"),
metricWithValueAndLabels(t, 1, "instance", "bar"),
},
map[uint64]*notifier.Alert{
hash(map[string]string{alertNameLabel: "common", "region": "east", "instance": "foo"}): {
Annotations: map[string]string{
"summary": `common: Too high connection number for "foo"`,
},
Labels: map[string]string{
alertNameLabel: "common",
"region": "east",
"instance": "foo",
},
},
hash(map[string]string{alertNameLabel: "common", "region": "east", "instance": "bar"}): {
Annotations: map[string]string{
"summary": `common: Too high connection number for "bar"`,
},
Labels: map[string]string{
alertNameLabel: "common",
"region": "east",
"instance": "bar",
},
},
},
},
{
&AlertingRule{
Name: "override label",
Labels: map[string]string{
"instance": "{{ $labels.instance }}",
},
Annotations: map[string]string{
"summary": `{{ $labels.__name__ }}: Too high connection number for "{{ $labels.instance }}"`,
"description": `{{ $labels.alertname}}: It is {{ $value }} connections for "{{ $labels.instance }}"`,
},
alerts: make(map[uint64]*notifier.Alert),
},
[]datasource.Metric{
metricWithValueAndLabels(t, 2, "__name__", "first", "instance", "foo", alertNameLabel, "override"),
metricWithValueAndLabels(t, 10, "__name__", "second", "instance", "bar", alertNameLabel, "override"),
},
map[uint64]*notifier.Alert{
hash(map[string]string{alertNameLabel: "override label", "exported_alertname": "override", "instance": "foo"}): {
Labels: map[string]string{
alertNameLabel: "override label",
"exported_alertname": "override",
"instance": "foo",
},
Annotations: map[string]string{
"summary": `first: Too high connection number for "foo"`,
"description": `override: It is 2 connections for "foo"`,
},
},
hash(map[string]string{alertNameLabel: "override label", "exported_alertname": "override", "instance": "bar"}): {
Labels: map[string]string{
alertNameLabel: "override label",
"exported_alertname": "override",
"instance": "bar",
},
Annotations: map[string]string{
"summary": `second: Too high connection number for "bar"`,
"description": `override: It is 10 connections for "bar"`,
},
},
},
},
{
&AlertingRule{
Name: "OriginLabels",
GroupName: "Testing",
Labels: map[string]string{
"instance": "{{ $labels.instance }}",
},
Annotations: map[string]string{
"summary": `Alert "{{ $labels.alertname }}({{ $labels.alertgroup }})" for instance {{ $labels.instance }}`,
},
alerts: make(map[uint64]*notifier.Alert),
},
[]datasource.Metric{
metricWithValueAndLabels(t, 1,
alertNameLabel, "originAlertname",
alertGroupNameLabel, "originGroupname",
"instance", "foo"),
},
map[uint64]*notifier.Alert{
hash(map[string]string{
alertNameLabel: "OriginLabels",
"exported_alertname": "originAlertname",
alertGroupNameLabel: "Testing",
"exported_alertgroup": "originGroupname",
"instance": "foo",
}): {
Labels: map[string]string{
alertNameLabel: "OriginLabels",
"exported_alertname": "originAlertname",
alertGroupNameLabel: "Testing",
"exported_alertgroup": "originGroupname",
"instance": "foo",
},
Annotations: map[string]string{
"summary": `Alert "originAlertname(originGroupname)" for instance foo`,
},
},
},
},
}
fakeGroup := Group{Name: "TestRule_Exec"}
for _, tc := range testCases {
t.Run(tc.rule.Name, func(t *testing.T) {
fq := &datasource.FakeQuerier{}
tc.rule.GroupID = fakeGroup.ID()
tc.rule.q = fq
tc.rule.state = &ruleState{entries: make([]StateEntry, 10)}
fq.Add(tc.metrics...)
if _, err := tc.rule.exec(context.TODO(), time.Now(), 0); err != nil {
t.Fatalf("unexpected err: %s", err)
}
for hash, expAlert := range tc.expAlerts {
gotAlert := tc.rule.alerts[hash]
if gotAlert == nil {
t.Fatalf("alert %d is missing; labels: %v; annotations: %v", hash, expAlert.Labels, expAlert.Annotations)
}
if !reflect.DeepEqual(expAlert.Annotations, gotAlert.Annotations) {
t.Fatalf("expected to have annotations %#v; got %#v", expAlert.Annotations, gotAlert.Annotations)
}
if !reflect.DeepEqual(expAlert.Labels, gotAlert.Labels) {
t.Fatalf("expected to have labels %#v; got %#v", expAlert.Labels, gotAlert.Labels)
}
}
})
}
}
func TestAlertsToSend(t *testing.T) {
ts := time.Now()
f := func(alerts, expAlerts []*notifier.Alert, resolveDuration, resendDelay time.Duration) {
t.Helper()
ar := &AlertingRule{alerts: make(map[uint64]*notifier.Alert)}
for i, a := range alerts {
ar.alerts[uint64(i)] = a
}
gotAlerts := ar.alertsToSend(ts, resolveDuration, resendDelay)
if gotAlerts == nil && expAlerts == nil {
return
}
if len(gotAlerts) != len(expAlerts) {
t.Fatalf("expected to get %d alerts; got %d instead",
len(expAlerts), len(gotAlerts))
}
sort.Slice(expAlerts, func(i, j int) bool {
return expAlerts[i].Name < expAlerts[j].Name
})
sort.Slice(gotAlerts, func(i, j int) bool {
return gotAlerts[i].Name < gotAlerts[j].Name
})
for i, exp := range expAlerts {
got := gotAlerts[i]
if got.LastSent != exp.LastSent {
t.Fatalf("expected LastSent to be %v; got %v", exp.LastSent, got.LastSent)
}
if got.End != exp.End {
t.Fatalf("expected End to be %v; got %v", exp.End, got.End)
}
}
}
f( // send firing alert with custom resolve time
[]*notifier.Alert{{State: notifier.StateFiring}},
[]*notifier.Alert{{LastSent: ts, End: ts.Add(5 * time.Minute)}},
5*time.Minute, time.Minute,
)
f( // resolve inactive alert at the current timestamp
Vmalert compliance 2 (#2340) * vmalert: split alert's `Start` field into `ActiveAt` and `Start` The `ActiveAt` field identifies when alert becomes active for rules with `for > 0`. Previously, this value was stored in field `Start`. The field `Start` now identifies the moment alert became `FIRING`. The split is needed in order to distinguish these two moments in the API responses for alerts. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: support specific moment of time for rules evaluation The Querier interface was extended to accept a new argument used as a timestamp at which evaluation should be made. It is needed to align rules execution time within the group. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: mark disappeared series as stale Series generated by alerting rules, which were sent to remote write now will be marked as stale if they will disappear on the next evaluation. This would make ALERTS and ALERTS_FOR_TIME series more precise. Signed-off-by: hagen1778 <roman@victoriametrics.com> * wip Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: evaluate rules at fixed timestamp Before, time at which rules were evaluated was calculated right before rule execution. The change makes sure that timestamp is calculated only once per evalution round and all rules are using the same timestamp. It also updates the logic of resending of already resolved alert notification. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: allow overridin `alertname` label value if it is present in response Previously, `alertname` was always equal to the Alerting Rule name. Now, its value can be overriden if series in response containt the different value for this label. The change is needed for improving compatibility with Prometheus. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: align rules evaluation in time Now, evaluation timestamp for rules evaluates as if there was no delay in rules evaluation. It means, that rules will be evaluated at fixed timestamps+group_interval. This way provides more consistent evaluation results and improves compatibility with Prometheus, Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: add metric for missed iterations New metric `vmalert_iteration_missed_total` will show whether rules evaluation round was missed. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: reduce delay before the initial rule evaluation in group Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: rollback alertname override According to the spec: ``` The alert name from the alerting rule (HighRequestLatency from the example above) MUST be added to the labels of the alert with the label name as alertname. It MUST override any existing alertname label. ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-3 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: throw err immediately on dedup detection ``` The execution of an alerting rule MUST error out immediately and MUST NOT send any alerts or add samples to samples receiver if there is more than one alert with the same labels ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-4 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: cleanup Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: use strings builder to reduce allocs Signed-off-by: hagen1778 <roman@victoriametrics.com>
2022-03-29 15:09:07 +02:00
[]*notifier.Alert{{State: notifier.StateInactive, ResolvedAt: ts}},
[]*notifier.Alert{{LastSent: ts, End: ts}},
time.Minute, time.Minute,
)
f( // mixed case of firing and resolved alerts. Names are added for deterministic sorting
Vmalert compliance 2 (#2340) * vmalert: split alert's `Start` field into `ActiveAt` and `Start` The `ActiveAt` field identifies when alert becomes active for rules with `for > 0`. Previously, this value was stored in field `Start`. The field `Start` now identifies the moment alert became `FIRING`. The split is needed in order to distinguish these two moments in the API responses for alerts. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: support specific moment of time for rules evaluation The Querier interface was extended to accept a new argument used as a timestamp at which evaluation should be made. It is needed to align rules execution time within the group. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: mark disappeared series as stale Series generated by alerting rules, which were sent to remote write now will be marked as stale if they will disappear on the next evaluation. This would make ALERTS and ALERTS_FOR_TIME series more precise. Signed-off-by: hagen1778 <roman@victoriametrics.com> * wip Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: evaluate rules at fixed timestamp Before, time at which rules were evaluated was calculated right before rule execution. The change makes sure that timestamp is calculated only once per evalution round and all rules are using the same timestamp. It also updates the logic of resending of already resolved alert notification. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: allow overridin `alertname` label value if it is present in response Previously, `alertname` was always equal to the Alerting Rule name. Now, its value can be overriden if series in response containt the different value for this label. The change is needed for improving compatibility with Prometheus. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: align rules evaluation in time Now, evaluation timestamp for rules evaluates as if there was no delay in rules evaluation. It means, that rules will be evaluated at fixed timestamps+group_interval. This way provides more consistent evaluation results and improves compatibility with Prometheus, Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: add metric for missed iterations New metric `vmalert_iteration_missed_total` will show whether rules evaluation round was missed. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: reduce delay before the initial rule evaluation in group Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: rollback alertname override According to the spec: ``` The alert name from the alerting rule (HighRequestLatency from the example above) MUST be added to the labels of the alert with the label name as alertname. It MUST override any existing alertname label. ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-3 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: throw err immediately on dedup detection ``` The execution of an alerting rule MUST error out immediately and MUST NOT send any alerts or add samples to samples receiver if there is more than one alert with the same labels ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-4 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: cleanup Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: use strings builder to reduce allocs Signed-off-by: hagen1778 <roman@victoriametrics.com>
2022-03-29 15:09:07 +02:00
[]*notifier.Alert{{Name: "a", State: notifier.StateFiring}, {Name: "b", State: notifier.StateInactive, ResolvedAt: ts}},
[]*notifier.Alert{{Name: "a", LastSent: ts, End: ts.Add(5 * time.Minute)}, {Name: "b", LastSent: ts, End: ts}},
5*time.Minute, time.Minute,
)
f( // mixed case of pending and resolved alerts. Names are added for deterministic sorting
Vmalert compliance 2 (#2340) * vmalert: split alert's `Start` field into `ActiveAt` and `Start` The `ActiveAt` field identifies when alert becomes active for rules with `for > 0`. Previously, this value was stored in field `Start`. The field `Start` now identifies the moment alert became `FIRING`. The split is needed in order to distinguish these two moments in the API responses for alerts. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: support specific moment of time for rules evaluation The Querier interface was extended to accept a new argument used as a timestamp at which evaluation should be made. It is needed to align rules execution time within the group. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: mark disappeared series as stale Series generated by alerting rules, which were sent to remote write now will be marked as stale if they will disappear on the next evaluation. This would make ALERTS and ALERTS_FOR_TIME series more precise. Signed-off-by: hagen1778 <roman@victoriametrics.com> * wip Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: evaluate rules at fixed timestamp Before, time at which rules were evaluated was calculated right before rule execution. The change makes sure that timestamp is calculated only once per evalution round and all rules are using the same timestamp. It also updates the logic of resending of already resolved alert notification. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: allow overridin `alertname` label value if it is present in response Previously, `alertname` was always equal to the Alerting Rule name. Now, its value can be overriden if series in response containt the different value for this label. The change is needed for improving compatibility with Prometheus. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: align rules evaluation in time Now, evaluation timestamp for rules evaluates as if there was no delay in rules evaluation. It means, that rules will be evaluated at fixed timestamps+group_interval. This way provides more consistent evaluation results and improves compatibility with Prometheus, Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: add metric for missed iterations New metric `vmalert_iteration_missed_total` will show whether rules evaluation round was missed. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: reduce delay before the initial rule evaluation in group Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: rollback alertname override According to the spec: ``` The alert name from the alerting rule (HighRequestLatency from the example above) MUST be added to the labels of the alert with the label name as alertname. It MUST override any existing alertname label. ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-3 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: throw err immediately on dedup detection ``` The execution of an alerting rule MUST error out immediately and MUST NOT send any alerts or add samples to samples receiver if there is more than one alert with the same labels ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-4 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: cleanup Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: use strings builder to reduce allocs Signed-off-by: hagen1778 <roman@victoriametrics.com>
2022-03-29 15:09:07 +02:00
[]*notifier.Alert{{Name: "a", State: notifier.StatePending}, {Name: "b", State: notifier.StateInactive, ResolvedAt: ts}},
[]*notifier.Alert{{Name: "b", LastSent: ts, End: ts}},
5*time.Minute, time.Minute,
)
f( // attempt to send alert that was already sent in the resendDelay interval
[]*notifier.Alert{{State: notifier.StateFiring, LastSent: ts.Add(-time.Second)}},
nil,
time.Minute, time.Minute,
)
f( // attempt to send alert that was sent out of the resendDelay interval
[]*notifier.Alert{{State: notifier.StateFiring, LastSent: ts.Add(-2 * time.Minute)}},
[]*notifier.Alert{{LastSent: ts, End: ts.Add(time.Minute)}},
time.Minute, time.Minute,
)
f( // alert must be sent even if resendDelay interval is 0
[]*notifier.Alert{{State: notifier.StateFiring, LastSent: ts.Add(-time.Second)}},
[]*notifier.Alert{{LastSent: ts, End: ts.Add(time.Minute)}},
time.Minute, 0,
)
Vmalert compliance 2 (#2340) * vmalert: split alert's `Start` field into `ActiveAt` and `Start` The `ActiveAt` field identifies when alert becomes active for rules with `for > 0`. Previously, this value was stored in field `Start`. The field `Start` now identifies the moment alert became `FIRING`. The split is needed in order to distinguish these two moments in the API responses for alerts. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: support specific moment of time for rules evaluation The Querier interface was extended to accept a new argument used as a timestamp at which evaluation should be made. It is needed to align rules execution time within the group. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: mark disappeared series as stale Series generated by alerting rules, which were sent to remote write now will be marked as stale if they will disappear on the next evaluation. This would make ALERTS and ALERTS_FOR_TIME series more precise. Signed-off-by: hagen1778 <roman@victoriametrics.com> * wip Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: evaluate rules at fixed timestamp Before, time at which rules were evaluated was calculated right before rule execution. The change makes sure that timestamp is calculated only once per evalution round and all rules are using the same timestamp. It also updates the logic of resending of already resolved alert notification. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: allow overridin `alertname` label value if it is present in response Previously, `alertname` was always equal to the Alerting Rule name. Now, its value can be overriden if series in response containt the different value for this label. The change is needed for improving compatibility with Prometheus. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: align rules evaluation in time Now, evaluation timestamp for rules evaluates as if there was no delay in rules evaluation. It means, that rules will be evaluated at fixed timestamps+group_interval. This way provides more consistent evaluation results and improves compatibility with Prometheus, Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: add metric for missed iterations New metric `vmalert_iteration_missed_total` will show whether rules evaluation round was missed. Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: reduce delay before the initial rule evaluation in group Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: rollback alertname override According to the spec: ``` The alert name from the alerting rule (HighRequestLatency from the example above) MUST be added to the labels of the alert with the label name as alertname. It MUST override any existing alertname label. ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-3 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: throw err immediately on dedup detection ``` The execution of an alerting rule MUST error out immediately and MUST NOT send any alerts or add samples to samples receiver if there is more than one alert with the same labels ``` https://github.com/prometheus/compliance/blob/main/alert_generator/specification.md#step-4 Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: cleanup Signed-off-by: hagen1778 <roman@victoriametrics.com> * vmalert: use strings builder to reduce allocs Signed-off-by: hagen1778 <roman@victoriametrics.com>
2022-03-29 15:09:07 +02:00
f( // inactive alert which has been sent already
[]*notifier.Alert{{State: notifier.StateInactive, LastSent: ts.Add(-time.Second), ResolvedAt: ts.Add(-2 * time.Second)}},
nil,
time.Minute, time.Minute,
)
f( // inactive alert which has been resolved after last send
[]*notifier.Alert{{State: notifier.StateInactive, LastSent: ts.Add(-time.Second), ResolvedAt: ts}},
[]*notifier.Alert{{LastSent: ts, End: ts}},
time.Minute, time.Minute,
)
}
func newTestRuleWithLabels(name string, labels ...string) *AlertingRule {
r := newTestAlertingRule(name, 0)
r.Labels = make(map[string]string)
for i := 0; i < len(labels); i += 2 {
r.Labels[labels[i]] = labels[i+1]
}
return r
}
func newTestAlertingRule(name string, waitFor time.Duration) *AlertingRule {
rule := AlertingRule{
Name: name,
For: waitFor,
EvalInterval: waitFor,
alerts: make(map[uint64]*notifier.Alert),
state: &ruleState{entries: make([]StateEntry, 10)},
metrics: &alertingRuleMetrics{
errors: utils.GetOrCreateCounter(fmt.Sprintf(`vmalert_alerting_rules_errors_total{alertname=%q}`, name)),
},
}
return &rule
}
func newTestAlertingRuleWithKeepFiring(name string, waitFor, keepFiringFor time.Duration) *AlertingRule {
rule := newTestAlertingRule(name, waitFor)
rule.KeepFiringFor = keepFiringFor
return rule
}
func TestAlertingRule_ToLabels(t *testing.T) {
metric := datasource.Metric{
Labels: []datasource.Label{
{Name: "instance", Value: "0.0.0.0:8800"},
{Name: "group", Value: "vmalert"},
{Name: "alertname", Value: "ConfigurationReloadFailure"},
},
Values: []float64{1},
Timestamps: []int64{time.Now().UnixNano()},
}
ar := &AlertingRule{
Labels: map[string]string{
"instance": "override", // this should override instance with new value
"group": "vmalert", // this shouldn't have effect since value in metric is equal
},
Expr: "sum(vmalert_alerting_rules_error) by(instance, group, alertname) > 0",
Name: "AlertingRulesError",
GroupName: "vmalert",
}
expectedOriginLabels := map[string]string{
"instance": "0.0.0.0:8800",
"group": "vmalert",
"alertname": "ConfigurationReloadFailure",
"alertgroup": "vmalert",
}
expectedProcessedLabels := map[string]string{
"instance": "override",
"exported_instance": "0.0.0.0:8800",
"alertname": "AlertingRulesError",
"exported_alertname": "ConfigurationReloadFailure",
"group": "vmalert",
"alertgroup": "vmalert",
}
ls, err := ar.toLabels(metric, nil)
if err != nil {
t.Fatalf("unexpected error: %s", err)
}
if !reflect.DeepEqual(ls.origin, expectedOriginLabels) {
t.Errorf("origin labels mismatch, got: %v, want: %v", ls.origin, expectedOriginLabels)
}
if !reflect.DeepEqual(ls.processed, expectedProcessedLabels) {
t.Errorf("processed labels mismatch, got: %v, want: %v", ls.processed, expectedProcessedLabels)
}
}