k8s.io/perf-tests/clusterloader2@v0.0.0-20240304094227-64bdb12da87e/pkg/measurement/common/nodelocaldns_latency_prometheus.go (about)

     1  /*
     2  Copyright 2021 The Kubernetes Authors.
     3  
     4  Licensed under the Apache License, Version 2.0 (the "License");
     5  you may not use this file except in compliance with the License.
     6  You may obtain a copy of the License at
     7  
     8      http://www.apache.org/licenses/LICENSE-2.0
     9  
    10  Unless required by applicable law or agreed to in writing, software
    11  distributed under the License is distributed on an "AS IS" BASIS,
    12  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
    13  See the License for the specific language governing permissions and
    14  limitations under the License.
    15  */
    16  
    17  package common
    18  
    19  import (
    20  	"fmt"
    21  	"time"
    22  
    23  	"k8s.io/klog/v2"
    24  	"k8s.io/perf-tests/clusterloader2/pkg/errors"
    25  	"k8s.io/perf-tests/clusterloader2/pkg/measurement"
    26  	measurementutil "k8s.io/perf-tests/clusterloader2/pkg/measurement/util"
    27  	"k8s.io/perf-tests/clusterloader2/pkg/util"
    28  )
    29  
    30  const (
    31  	nodelocaldnsLatencyPrometheusMeasurementName = "NodeLocalDNSLatencyPrometheus"
    32  	percLatencyQueryTemplate                     = `histogram_quantile(%v, sum(rate(coredns_dns_request_duration_seconds_bucket[%v])) by (le))`
    33  	defaultThreshold                             = 5 * time.Second
    34  )
    35  
    36  var (
    37  	desiredPercentiles = []float64{0.5, 0.9, 0.99}
    38  )
    39  
    40  func init() {
    41  	create := func() measurement.Measurement {
    42  		return CreatePrometheusMeasurement(&nodelocaldnsLatencyGatherer{})
    43  	}
    44  	if err := measurement.Register(nodelocaldnsLatencyPrometheusMeasurementName, create); err != nil {
    45  		klog.Fatalf("Cannot register %s: %v", nodelocaldnsLatencyPrometheusMeasurementName, err)
    46  	}
    47  }
    48  
    49  type nodelocaldnsLatencyGatherer struct{}
    50  
    51  func (n *nodelocaldnsLatencyGatherer) Gather(executor QueryExecutor, startTime, endTime time.Time, config *measurement.Config) ([]measurement.Summary, error) {
    52  	result, err := n.getPercentileLatencies(executor, startTime, endTime)
    53  	if err != nil {
    54  		return nil, err
    55  	}
    56  	content, err := util.PrettyPrintJSON(result.ToPerfData(n.String()))
    57  	if err != nil {
    58  		return nil, err
    59  	}
    60  	summaries := []measurement.Summary{measurement.CreateSummary(n.String(), "json", content)}
    61  	return summaries, n.validateResult(config, result)
    62  }
    63  
    64  func (n *nodelocaldnsLatencyGatherer) String() string {
    65  	return nodelocaldnsLatencyPrometheusMeasurementName
    66  }
    67  
    68  func (n *nodelocaldnsLatencyGatherer) Configure(config *measurement.Config) error {
    69  	return nil
    70  }
    71  func (n *nodelocaldnsLatencyGatherer) IsEnabled(config *measurement.Config) bool {
    72  	return true
    73  }
    74  
    75  func (n *nodelocaldnsLatencyGatherer) validateResult(config *measurement.Config, result *measurementutil.LatencyMetric) error {
    76  	latencyUpperBound, err := util.GetDurationOrDefault(config.Params, "threshold", defaultThreshold)
    77  	if err != nil {
    78  		return err
    79  	}
    80  	if result.Perc99 > latencyUpperBound {
    81  		return errors.NewMetricViolationError(
    82  			"NodelocalDNS dns_request_duration_seconds",
    83  			fmt.Sprintf("99th Percentile Latency %v is higher than the upper bound of %s", result.Perc99, latencyUpperBound))
    84  	}
    85  	return nil
    86  }
    87  
    88  func (n *nodelocaldnsLatencyGatherer) getPercentileLatencies(executor QueryExecutor, startTime, endTime time.Time) (*measurementutil.LatencyMetric, error) {
    89  	measurementDuration := endTime.Sub(startTime)
    90  	promDuration := measurementutil.ToPrometheusTime(measurementDuration)
    91  	errList := errors.NewErrorList()
    92  	result := &measurementutil.LatencyMetric{}
    93  	for _, percVal := range desiredPercentiles {
    94  		query := fmt.Sprintf(percLatencyQueryTemplate, percVal, promDuration)
    95  		samples, err := executor.Query(query, endTime)
    96  		if err != nil {
    97  			errList.Append(fmt.Errorf("failed to execute query %q, err - %v", query, err))
    98  			continue
    99  		}
   100  		if len(samples) != 1 {
   101  			errList.Append(fmt.Errorf("got unexpected number of samples: %d for query %q", len(samples), query))
   102  			continue
   103  		}
   104  		result.SetQuantile(percVal, time.Duration(float64(samples[0].Value)*float64(time.Second)))
   105  	}
   106  	if !errList.IsEmpty() {
   107  		return nil, fmt.Errorf("failed to compute latencies, errors - %s", errList.Error())
   108  	}
   109  	return result, nil
   110  }