VictoriaMetrics/app/vmalert-tool/unittest/recording.go
Hui Wang b09272ccac
Some checks are pending
build / Build (push) Waiting to run
CodeQL Go / Analyze (push) Waiting to run
main / lint (push) Waiting to run
main / test (test-full) (push) Blocked by required conditions
main / test (test-full-386) (push) Blocked by required conditions
main / test (test-pure) (push) Blocked by required conditions
publish-docs / Build (push) Waiting to run
app/vmalert: improve performances when rules produce large volumes of results
1. Avoid storing the last evaluation results outside of rules, check for
stale time series as soon as possible;
2. remove duplicated template `Clone()`.

This pull request is primarily reducing memory usage when rules produce
large volumes of results, as seen in
https://github.com/VictoriaMetrics/VictoriaMetrics/issues/6894.
The CPU time spent on garbage collection remains high and may be
addressed in a separate PR.
2024-11-14 12:23:39 +01:00

104 lines
3.4 KiB
Go

package unittest
import (
"context"
"fmt"
"net/url"
"reflect"
"sort"
"time"
"github.com/VictoriaMetrics/VictoriaMetrics/app/vmalert/datasource"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/prompbmarshal"
"github.com/VictoriaMetrics/VictoriaMetrics/lib/promutils"
"github.com/VictoriaMetrics/metricsql"
)
// metricsqlTestCase holds metricsql_expr_test cases defined in test file
type metricsqlTestCase struct {
Expr string `yaml:"expr"`
EvalTime *promutils.Duration `yaml:"eval_time"`
ExpSamples []expSample `yaml:"exp_samples"`
}
type expSample struct {
Labels string `yaml:"labels"`
Value float64 `yaml:"value"`
}
// checkMetricsqlCase will check metricsql_expr_test cases
func checkMetricsqlCase(cases []metricsqlTestCase, q datasource.QuerierBuilder) (checkErrs []error) {
queries := q.BuildWithParams(datasource.QuerierParams{QueryParams: url.Values{"nocache": {"1"}, "latency_offset": {"1ms"}}, DataSourceType: "prometheus"})
Outer:
for _, mt := range cases {
result, _, err := queries.Query(context.Background(), mt.Expr, durationToTime(mt.EvalTime))
if err != nil {
checkErrs = append(checkErrs, fmt.Errorf(" expr: %q, time: %s, err: %w", mt.Expr,
mt.EvalTime.Duration().String(), err))
continue
}
var gotSamples []parsedSample
for _, s := range result.Data {
sort.Slice(s.Labels, func(i, j int) bool {
return s.Labels[i].Name < s.Labels[j].Name
})
gotSamples = append(gotSamples, parsedSample{
Labels: s.Labels,
Value: s.Values[0],
})
}
var expSamples []parsedSample
for _, s := range mt.ExpSamples {
expLb := []prompbmarshal.Label{}
if s.Labels != "" {
metricsqlExpr, err := metricsql.Parse(s.Labels)
if err != nil {
checkErrs = append(checkErrs, fmt.Errorf("\n expr: %q, time: %s, err: %v", mt.Expr,
mt.EvalTime.Duration().String(), fmt.Errorf("failed to parse labels %q: %w", s.Labels, err)))
continue Outer
}
metricsqlMetricExpr, ok := metricsqlExpr.(*metricsql.MetricExpr)
if !ok || len(metricsqlMetricExpr.LabelFilterss) > 1 {
checkErrs = append(checkErrs, fmt.Errorf("\n expr: %q, time: %s, err: %v", mt.Expr,
mt.EvalTime.Duration().String(), fmt.Errorf("got invalid exp_samples: %q", s.Labels)))
continue Outer
}
if len(metricsqlMetricExpr.LabelFilterss) > 0 {
for _, l := range metricsqlMetricExpr.LabelFilterss[0] {
expLb = append(expLb, prompbmarshal.Label{
Name: l.Label,
Value: l.Value,
})
}
}
}
sort.Slice(expLb, func(i, j int) bool {
return expLb[i].Name < expLb[j].Name
})
expSamples = append(expSamples, parsedSample{
Labels: expLb,
Value: s.Value,
})
}
sort.Slice(expSamples, func(i, j int) bool {
return datasource.LabelCompare(expSamples[i].Labels, expSamples[j].Labels) <= 0
})
sort.Slice(gotSamples, func(i, j int) bool {
return datasource.LabelCompare(gotSamples[i].Labels, gotSamples[j].Labels) <= 0
})
if !reflect.DeepEqual(expSamples, gotSamples) {
checkErrs = append(checkErrs, fmt.Errorf("\n expr: %q, time: %s,\n exp: %v\n got: %v", mt.Expr,
mt.EvalTime.Duration().String(), parsedSamplesString(expSamples), parsedSamplesString(gotSamples)))
}
}
return
}
func durationToTime(pd *promutils.Duration) time.Time {
if pd == nil {
return time.Time{}
}
return time.UnixMilli(pd.Duration().Milliseconds())
}