美文网首页
聊聊springboot的http.server.request

聊聊springboot的http.server.request

作者: go4it | 来源:发表于2023-11-30 09:13 被阅读0次

    本文主要研究一下springboot的http.server.requests

    http.server.requests

    org/springframework/boot/actuate/autoconfigure/metrics/MetricsProperties.java

            public static class Server {
    
                private final ServerRequest request = new ServerRequest();
    
                /**
                 * Maximum number of unique URI tag values allowed. After the max number of
                 * tag values is reached, metrics with additional tag values are denied by
                 * filter.
                 */
                private int maxUriTags = 100;
    
                public ServerRequest getRequest() {
                    return this.request;
                }
    
                public int getMaxUriTags() {
                    return this.maxUriTags;
                }
    
                public void setMaxUriTags(int maxUriTags) {
                    this.maxUriTags = maxUriTags;
                }
    
                public static class ServerRequest {
    
                    /**
                     * Name of the metric for received requests.
                     */
                    private String metricName = "http.server.requests";
    
                    /**
                     * Whether the trailing slash should be ignored when recording metrics.
                     */
                    private boolean ignoreTrailingSlash = true;
    
                    //......
                }   
            }    
    

    MetricsProperties.Server.ServerRequest定义了http.server.requests这个metrics名

    WebMvcMetricsAutoConfiguration

    org/springframework/boot/actuate/autoconfigure/metrics/web/servlet/WebMvcMetricsAutoConfiguration.java

    @Configuration(proxyBeanMethods = false)
    @AutoConfigureAfter({ MetricsAutoConfiguration.class, CompositeMeterRegistryAutoConfiguration.class,
            SimpleMetricsExportAutoConfiguration.class })
    @ConditionalOnWebApplication(type = ConditionalOnWebApplication.Type.SERVLET)
    @ConditionalOnClass(DispatcherServlet.class)
    @ConditionalOnBean(MeterRegistry.class)
    @EnableConfigurationProperties(MetricsProperties.class)
    public class WebMvcMetricsAutoConfiguration {
    
        private final MetricsProperties properties;
    
        public WebMvcMetricsAutoConfiguration(MetricsProperties properties) {
            this.properties = properties;
        }
    
        @Bean
        @ConditionalOnMissingBean(WebMvcTagsProvider.class)
        public DefaultWebMvcTagsProvider webMvcTagsProvider(ObjectProvider<WebMvcTagsContributor> contributors) {
            return new DefaultWebMvcTagsProvider(this.properties.getWeb().getServer().getRequest().isIgnoreTrailingSlash(),
                    contributors.orderedStream().collect(Collectors.toList()));
        }
    
        @Bean
        public FilterRegistrationBean<WebMvcMetricsFilter> webMvcMetricsFilter(MeterRegistry registry,
                WebMvcTagsProvider tagsProvider) {
            ServerRequest request = this.properties.getWeb().getServer().getRequest();
            WebMvcMetricsFilter filter = new WebMvcMetricsFilter(registry, tagsProvider, request.getMetricName(),
                    request.getAutotime());
            FilterRegistrationBean<WebMvcMetricsFilter> registration = new FilterRegistrationBean<>(filter);
            registration.setOrder(Ordered.HIGHEST_PRECEDENCE + 1);
            registration.setDispatcherTypes(DispatcherType.REQUEST, DispatcherType.ASYNC);
            return registration;
        }
    
        @Bean
        @Order(0)
        public MeterFilter metricsHttpServerUriTagFilter() {
            String metricName = this.properties.getWeb().getServer().getRequest().getMetricName();
            MeterFilter filter = new OnlyOnceLoggingDenyMeterFilter(
                    () -> String.format("Reached the maximum number of URI tags for '%s'.", metricName));
            return MeterFilter.maximumAllowableTags(metricName, "uri", this.properties.getWeb().getServer().getMaxUriTags(),
                    filter);
        }
    
        @Bean
        public MetricsWebMvcConfigurer metricsWebMvcConfigurer(MeterRegistry meterRegistry,
                WebMvcTagsProvider tagsProvider) {
            return new MetricsWebMvcConfigurer(meterRegistry, tagsProvider);
        }
    }    
    

    WebMvcMetricsAutoConfiguration定义了webMvcTagsProvider,可以接收WebMvcTagsContributor对tag进行定制,还注册了WebMvcMetricsFilter、MeterFilter、metricsWebMvcConfigurer

    DefaultWebMvcTagsProvider

    org/springframework/boot/actuate/metrics/web/servlet/DefaultWebMvcTagsProvider.java

    public class DefaultWebMvcTagsProvider implements WebMvcTagsProvider {
    
        private final boolean ignoreTrailingSlash;
    
        private final List<WebMvcTagsContributor> contributors;
    
        public DefaultWebMvcTagsProvider() {
            this(false);
        }
    
        /**
         * Creates a new {@link DefaultWebMvcTagsProvider} that will provide tags from the
         * given {@code contributors} in addition to its own.
         * @param contributors the contributors that will provide additional tags
         * @since 2.3.0
         */
        public DefaultWebMvcTagsProvider(List<WebMvcTagsContributor> contributors) {
            this(false, contributors);
        }
    
        public DefaultWebMvcTagsProvider(boolean ignoreTrailingSlash) {
            this(ignoreTrailingSlash, Collections.emptyList());
        }
    
        /**
         * Creates a new {@link DefaultWebMvcTagsProvider} that will provide tags from the
         * given {@code contributors} in addition to its own.
         * @param ignoreTrailingSlash whether trailing slashes should be ignored when
         * determining the {@code uri} tag.
         * @param contributors the contributors that will provide additional tags
         * @since 2.3.0
         */
        public DefaultWebMvcTagsProvider(boolean ignoreTrailingSlash, List<WebMvcTagsContributor> contributors) {
            this.ignoreTrailingSlash = ignoreTrailingSlash;
            this.contributors = contributors;
        }
    
        @Override
        public Iterable<Tag> getTags(HttpServletRequest request, HttpServletResponse response, Object handler,
                Throwable exception) {
            Tags tags = Tags.of(WebMvcTags.method(request), WebMvcTags.uri(request, response, this.ignoreTrailingSlash),
                    WebMvcTags.exception(exception), WebMvcTags.status(response), WebMvcTags.outcome(response));
            for (WebMvcTagsContributor contributor : this.contributors) {
                tags = tags.and(contributor.getTags(request, response, handler, exception));
            }
            return tags;
        }
    
        @Override
        public Iterable<Tag> getLongRequestTags(HttpServletRequest request, Object handler) {
            Tags tags = Tags.of(WebMvcTags.method(request), WebMvcTags.uri(request, null, this.ignoreTrailingSlash));
            for (WebMvcTagsContributor contributor : this.contributors) {
                tags = tags.and(contributor.getLongRequestTags(request, handler));
            }
            return tags;
        }
    
    }
    

    DefaultWebMvcTagsProvider实现了WebMvcTagsProvider接口,其构造器可以接收WebMvcTagsContributor,对tag进行定制;默认的tag为WebMvcTags.method(method)、WebMvcTags.uri(uri)、WebMvcTags.exception(exception)、WebMvcTags.status(status)、WebMvcTags.outcome(outcome)

    WebMvcMetricsFilter

    org/springframework/boot/actuate/metrics/web/servlet/WebMvcMetricsFilter.java

    public class WebMvcMetricsFilter extends OncePerRequestFilter {
    
        private final MeterRegistry registry;
    
        private final WebMvcTagsProvider tagsProvider;
    
        private final String metricName;
    
        private final AutoTimer autoTimer;
    
        /**
         * Create a new {@link WebMvcMetricsFilter} instance.
         * @param registry the meter registry
         * @param tagsProvider the tags provider
         * @param metricName the metric name
         * @param autoTimer the auto-timers to apply or {@code null} to disable auto-timing
         * @since 2.2.0
         */
        public WebMvcMetricsFilter(MeterRegistry registry, WebMvcTagsProvider tagsProvider, String metricName,
                AutoTimer autoTimer) {
            this.registry = registry;
            this.tagsProvider = tagsProvider;
            this.metricName = metricName;
            this.autoTimer = autoTimer;
        }
    
        @Override
        protected boolean shouldNotFilterAsyncDispatch() {
            return false;
        }
    
        @Override
        protected void doFilterInternal(HttpServletRequest request, HttpServletResponse response, FilterChain filterChain)
                throws ServletException, IOException {
            TimingContext timingContext = TimingContext.get(request);
            if (timingContext == null) {
                timingContext = startAndAttachTimingContext(request);
            }
            try {
                filterChain.doFilter(request, response);
                if (!request.isAsyncStarted()) {
                    // Only record when async processing has finished or never been started.
                    // If async was started by something further down the chain we wait
                    // until the second filter invocation (but we'll be using the
                    // TimingContext that was attached to the first)
                    Throwable exception = (Throwable) request.getAttribute(DispatcherServlet.EXCEPTION_ATTRIBUTE);
                    record(timingContext, request, response, exception);
                }
            }
            catch (NestedServletException ex) {
                response.setStatus(HttpStatus.INTERNAL_SERVER_ERROR.value());
                record(timingContext, request, response, ex.getCause());
                throw ex;
            }
            catch (ServletException | IOException | RuntimeException ex) {
                record(timingContext, request, response, ex);
                throw ex;
            }
        }
    
        //......
    }    
    

    WebMvcMetricsFilter继承了OncePerRequestFilter,其doFilterInternal方法通过TimingContext来维护timerSample,然后在filterChain.doFilter(request, response)之后进行record

    record

        private void record(TimingContext timingContext, HttpServletRequest request, HttpServletResponse response,
                Throwable exception) {
            Object handler = getHandler(request);
            Set<Timed> annotations = getTimedAnnotations(handler);
            Timer.Sample timerSample = timingContext.getTimerSample();
            if (annotations.isEmpty()) {
                if (this.autoTimer.isEnabled()) {
                    Builder builder = this.autoTimer.builder(this.metricName);
                    timerSample.stop(getTimer(builder, handler, request, response, exception));
                }
            }
            else {
                for (Timed annotation : annotations) {
                    Builder builder = Timer.builder(annotation, this.metricName);
                    timerSample.stop(getTimer(builder, handler, request, response, exception));
                }
            }
        }
    
        private Timer getTimer(Builder builder, Object handler, HttpServletRequest request, HttpServletResponse response,
                Throwable exception) {
            return builder.tags(this.tagsProvider.getTags(request, response, handler, exception)).register(this.registry);
        }    
    

    record方法主要执行timerSample.stop(getTimer(builder, handler, request, response, exception))

    timerSample

    micrometer-core-1.5.9.jar!/io/micrometer/core/instrument/Timer.class

        public static class Sample {
            private Tags tags = Tags.empty();
            private final long startTime;
            private final Clock clock;
    
            Sample(Clock clock) {
                this.clock = clock;
                this.startTime = clock.monotonicTime();
            }
    
            public long stop(Timer timer) {
                long durationNs = this.clock.monotonicTime() - this.startTime;
                timer.record(durationNs, TimeUnit.NANOSECONDS);
                return durationNs;
            }
    
            @Incubating(
                since = "1.4.0"
            )
            public long stop(MeterRegistry registry, Builder timerBuilder) {
                return this.stop(timerBuilder.tags((Iterable)this.tags).register(registry));
            }
    
            @Incubating(
                since = "1.4.0"
            )
            public Sample tags(String... tags) {
                return this.tags((Iterable)Tags.of(tags));
            }
    
            @Incubating(
                since = "1.4.0"
            )
            public Sample tags(Iterable<Tag> tags) {
                this.tags = this.tags.and(tags);
                return this;
            }
        }
    

    Timer.Sample的stop方法主要执行timer.record(durationNs, TimeUnit.NANOSECONDS)

    PrometheusTimer

    io/micrometer/prometheus/PrometheusTimer.java

    public class PrometheusTimer extends AbstractTimer {
        private static final CountAtBucket[] EMPTY_HISTOGRAM = new CountAtBucket[0];
    
        private final LongAdder count = new LongAdder();
        private final LongAdder totalTime = new LongAdder();
        private final TimeWindowMax max;
    
        private final HistogramFlavor histogramFlavor;
    
        @Nullable
        private final Histogram histogram;
    
        PrometheusTimer(Id id, Clock clock, DistributionStatisticConfig distributionStatisticConfig, PauseDetector pauseDetector, HistogramFlavor histogramFlavor) {
            super(id, clock,
                    DistributionStatisticConfig.builder()
                            .percentilesHistogram(false)
                            .serviceLevelObjectives()
                            .build()
                            .merge(distributionStatisticConfig),
                    pauseDetector, TimeUnit.SECONDS, false);
    
            this.histogramFlavor = histogramFlavor;
            this.max = new TimeWindowMax(clock, distributionStatisticConfig);
    
            if (distributionStatisticConfig.isPublishingHistogram()) {
                switch (histogramFlavor) {
                    case Prometheus:
                        histogram = new TimeWindowFixedBoundaryHistogram(clock, DistributionStatisticConfig.builder()
                                .expiry(Duration.ofDays(1825)) // effectively never roll over
                                .bufferLength(1)
                                .build()
                                .merge(distributionStatisticConfig), true);
                        break;
                    case VictoriaMetrics:
                        histogram = new FixedBoundaryVictoriaMetricsHistogram();
                        break;
                    default:
                        histogram = null;
                        break;
                }
            } else {
                histogram = null;
            }
        }
    
        @Override
        protected void recordNonNegative(long amount, TimeUnit unit) {
            count.increment();
            long nanoAmount = TimeUnit.NANOSECONDS.convert(amount, unit);
            totalTime.add(nanoAmount);
            max.record(nanoAmount, TimeUnit.NANOSECONDS);
    
            if (histogram != null)
                histogram.recordLong(TimeUnit.NANOSECONDS.convert(amount, unit));
        }
    
        @Override
        public long count() {
            return count.longValue();
        }
    
        @Override
        public double totalTime(TimeUnit unit) {
            return TimeUtils.nanosToUnit(totalTime.doubleValue(), unit);
        }
    
        @Override
        public double max(TimeUnit unit) {
            return max.poll(unit);
        }
    
        public HistogramFlavor histogramFlavor() {
            return histogramFlavor;
        }
    
        /**
         * For Prometheus we cannot use the histogram counts from HistogramSnapshot, as it is based on a
         * rolling histogram. Prometheus requires a histogram that accumulates values over the lifetime of the app.
         *
         * @return Cumulative histogram buckets.
         */
        public CountAtBucket[] histogramCounts() {
            return histogram == null ? EMPTY_HISTOGRAM : histogram.takeSnapshot(0, 0, 0).histogramCounts();
        }
    
        @Override
        public HistogramSnapshot takeSnapshot() {
            HistogramSnapshot snapshot = super.takeSnapshot();
    
            if (histogram == null) {
                return snapshot;
            }
    
            return new HistogramSnapshot(snapshot.count(),
                    snapshot.total(),
                    snapshot.max(),
                    snapshot.percentileValues(),
                    histogramCounts(),
                    snapshot::outputSummary);
        }
    }
    

    PrometheusTimer继承了AbstractTimer,默认histogram为null

    PrometheusMeterRegistry

    io/micrometer/prometheus/PrometheusMeterRegistry.java

        protected io.micrometer.core.instrument.Timer newTimer(Meter.Id id, DistributionStatisticConfig distributionStatisticConfig, PauseDetector pauseDetector) {
            PrometheusTimer timer = new PrometheusTimer(id, clock, distributionStatisticConfig, pauseDetector, prometheusConfig.histogramFlavor());
            applyToCollector(id, (collector) ->
                    addDistributionStatisticSamples(distributionStatisticConfig, collector, timer, tagValues(id), false));
            return timer;
        }
    

    PrometheusMeterRegistry的newTimer创建的是PrometheusTimer,同时applyToCollector调用了addDistributionStatisticSamples

    addDistributionStatisticSamples

        private void addDistributionStatisticSamples(DistributionStatisticConfig distributionStatisticConfig, MicrometerCollector collector,
                                                     HistogramSupport histogramSupport, List<String> tagValues, boolean forLongTaskTimer) {
            collector.add(tagValues, (conventionName, tagKeys) -> {
                Stream.Builder<Collector.MetricFamilySamples.Sample> samples = Stream.builder();
    
                HistogramSnapshot histogramSnapshot = histogramSupport.takeSnapshot();
                ValueAtPercentile[] percentileValues = histogramSnapshot.percentileValues();
                CountAtBucket[] histogramCounts = histogramSnapshot.histogramCounts();
                double count = histogramSnapshot.count();
    
                if (percentileValues.length > 0) {
                    List<String> quantileKeys = new LinkedList<>(tagKeys);
                    quantileKeys.add("quantile");
    
                    // satisfies https://prometheus.io/docs/concepts/metric_types/#summary
                    for (ValueAtPercentile v : percentileValues) {
                        List<String> quantileValues = new LinkedList<>(tagValues);
                        quantileValues.add(Collector.doubleToGoString(v.percentile()));
                        samples.add(new Collector.MetricFamilySamples.Sample(
                                conventionName, quantileKeys, quantileValues, v.value(TimeUnit.SECONDS)));
                    }
                }
    
                Collector.Type type = distributionStatisticConfig.isPublishingHistogram() ? Collector.Type.HISTOGRAM : Collector.Type.SUMMARY;
                if (histogramCounts.length > 0) {
                    // Prometheus doesn't balk at a metric being BOTH a histogram and a summary
                    type = Collector.Type.HISTOGRAM;
    
                    List<String> histogramKeys = new LinkedList<>(tagKeys);
    
                    String sampleName = conventionName + "_bucket";
                    switch (prometheusConfig.histogramFlavor()) {
                        case Prometheus:
                            histogramKeys.add("le");
    
                            // satisfies https://prometheus.io/docs/concepts/metric_types/#histogram
                            for (CountAtBucket c : histogramCounts) {
                                final List<String> histogramValues = new LinkedList<>(tagValues);
                                histogramValues.add(Collector.doubleToGoString(c.bucket(TimeUnit.SECONDS)));
                                samples.add(new Collector.MetricFamilySamples.Sample(
                                        sampleName, histogramKeys, histogramValues, c.count()));
                            }
    
                            // the +Inf bucket should always equal `count`
                            final List<String> histogramValues = new LinkedList<>(tagValues);
                            histogramValues.add("+Inf");
                            samples.add(new Collector.MetricFamilySamples.Sample(
                                    sampleName, histogramKeys, histogramValues, count));
                            break;
                        case VictoriaMetrics:
                            histogramKeys.add("vmrange");
    
                            for (CountAtBucket c : histogramCounts) {
                                final List<String> histogramValuesVM = new LinkedList<>(tagValues);
                                histogramValuesVM.add(FixedBoundaryVictoriaMetricsHistogram.getRangeTagValue(c.bucket()));
                                samples.add(new Collector.MetricFamilySamples.Sample(
                                        sampleName, histogramKeys, histogramValuesVM, c.count()));
                            }
                            break;
                        default:
                            break;
                    }
    
                }
    
                samples.add(new Collector.MetricFamilySamples.Sample(
                        conventionName + (forLongTaskTimer ? "_active_count" : "_count"), tagKeys, tagValues, count));
    
                samples.add(new Collector.MetricFamilySamples.Sample(
                        conventionName + (forLongTaskTimer ? "_duration_sum" : "_sum"), tagKeys, tagValues, histogramSnapshot.total(TimeUnit.SECONDS)));
    
                return Stream.of(new MicrometerCollector.Family(type, conventionName, samples.build()),
                        new MicrometerCollector.Family(Collector.Type.GAUGE, conventionName + "_max", Stream.of(
                                new Collector.MetricFamilySamples.Sample(conventionName + "_max", tagKeys, tagValues,
                                        histogramSnapshot.max(getBaseTimeUnit())))));
            });
        }
    

    addDistributionStatisticSamples会判断isPublishingHistogram,是则发布HISTOGRAM,否则发布SUMMARY(默认为否),之后给samples添加_count及_sum的sample,最后额外添加一个_max的gauge

    示例

    示例1

    Timer timer = meterRegistry.timer("abc", "a", "b");
    timer.record(Duration.of(100, ChronoUnit.SECONDS));
    

    最后产生的metrics如下

    # HELP abc_seconds_max  
    # TYPE abc_seconds_max gauge
    abc_seconds_max{a="b",} 0.0
    # HELP abc_seconds  
    # TYPE abc_seconds summary
    abc_seconds_count{a="b",} 1.0
    abc_seconds_sum{a="b",} 100.0
    

    示例2

    http.server.requests的prometheus指标

    # HELP http_server_requests_seconds_max
    # TYPE http_server_requests_seconds_max gauge
    http_server_requests_seconds_max{}
    # TYPE http_server_requests_seconds summary
    http_server_requests_seconds_count{}
    http_server_requests_seconds_sum{}
    

    grafana展示

    HTTP Server Requests Count

                {
                  "expr": "http_server_requests_seconds_count{instance=\"$instance\", application=\"$application\"}",
                  "format": "time_series",
                  "interval": "",
                  "intervalFactor": 1,
                  "legendFormat": "{{method}} [{{status}}] - {{uri}}",
                  "refId": "A"
                }
    

    HTTP Server Requests Sum

                {
                  "expr": "http_server_requests_seconds_sum{instance=\"$instance\", application=\"$application\"}",
                  "format": "time_series",
                  "interval": "",
                  "intervalFactor": 1,
                  "legendFormat": "{{method}} [{{status}}] - {{uri}}",
                  "refId": "A"
                }
    

    HTTP Server Requests Max

                {
                  "expr": "http_server_requests_seconds_max{instance=\"$instance\", application=\"$application\"}",
                  "format": "time_series",
                  "interval": "",
                  "intervalFactor": 1,
                  "legendFormat": "{{method}} [{{status}}] - {{uri}}",
                  "refId": "A"
                }
    

    Total Requests

            {
              "expr": "sum(increase(http_server_requests_seconds_count{application=\"$application\", instance=~\"$instance\"}[150s]))",
              "format": "time_series",
              "instant": false,
              "interval": "",
              "intervalFactor": 1,
              "legendFormat": "",
              "refId": "A"
            }
    

    Request Count

            {
              "expr": "irate(http_server_requests_seconds_count{instance=\"$instance\", application=\"$application\", uri!~\".*actuator.*\"}[5m])",
              "format": "time_series",
              "interval": "",
              "intervalFactor": 1,
              "legendFormat": "{{method}} [{{status}}] - {{uri}} -{{reqPath}}",
              "refId": "A"
            }
    

    Failed Requests

            {
              "expr": "sum(increase(http_server_requests_seconds_count{application=\"$application\", instance=~\"$instance\", uri=~\"$uri\", status!=\"200\"}[$__interval]))",
              "format": "time_series",
              "intervalFactor": 1,
              "refId": "A"
            }
    

    Req / sec

            {
              "expr": "sum(irate(http_server_requests_seconds_count{application=\"$application\", instance=~\"$instance\"}[1m]))",
              "format": "time_series",
              "interval": "",
              "intervalFactor": 1,
              "legendFormat": "",
              "refId": "A"
            }
    

    Requests per second

            {
              "expr": "rate(http_server_requests_seconds_count{application=\"$application\", instance=\"$instance\"}[1m])",
              "format": "time_series",
              "intervalFactor": 1,
              "legendFormat": "{{method}}-{{status}}-{{uri}}",
              "refId": "A"
            }
    

    Top 10 Most Used API endpoints

            {
              "expr": "topk(10, sum by(uri, method) (rate(http_server_requests_seconds_count{application=\"$application\", instance=~\"$instance\"}[1m])))",
              "format": "time_series",
              "instant": false,
              "interval": "",
              "intervalFactor": 1,
              "legendFormat": "",
              "refId": "A"
            }
    

    Error Rate

            {
              "expr": "sum(increase(http_server_requests_seconds_count{application=\"$application\", instance=~\"$instance\", uri=~\"$uri\", status!=\"200\"}[$__interval])) / sum(increase(http_server_requests_seconds_count{application=\"$application\", instance=~\"$instance\", uri=~\"$uri\"}[$__interval])) * 100",
              "format": "time_series",
              "intervalFactor": 1,
              "refId": "A"
            }
    

    Mean response time

            {
              "expr": "rate(http_server_requests_seconds_sum{application=\"$application\", instance=\"$instance\"}[1m])/rate(http_server_requests_seconds_count{application=\"$application\", instance=\"$instance\"}[1m])",
              "format": "time_series",
              "instant": false,
              "intervalFactor": 1,
              "legendFormat": "{{method}}-{{status}}-{{uri}}",
              "refId": "A"
            }
    

    Response Time

            {
              "expr": "irate(http_server_requests_seconds_sum{instance=\"$instance\", application=\"$application\", exception=\"None\", uri!~\".*actuator.*\"}[5m]) / irate(http_server_requests_seconds_count{instance=\"$instance\", application=\"$application\", exception=\"None\", uri!~\".*actuator.*\"}[5m])",
              "format": "time_series",
              "intervalFactor": 1,
              "legendFormat": "{{method}} [{{status}}] - {{uri}}",
              "refId": "A"
            }
    

    Response time of 50%, 75%, 90%, 95% of requests

            {
              "expr": "histogram_quantile(0.95, sum(rate(http_server_requests_seconds_bucket{application=\"$application\", instance=\"$instance\"}[1m])) by (le))",
              "format": "time_series",
              "instant": false,
              "interval": "",
              "intervalFactor": 1,
              "legendFormat": "95%",
              "refId": "A"
            },
            {
              "expr": "histogram_quantile(0.9, sum(rate(http_server_requests_seconds_bucket{application=\"$application\", instance=\"$instance\"}[1m])) by (le))",
              "format": "time_series",
              "intervalFactor": 1,
              "legendFormat": "90%",
              "refId": "B"
            },
            {
              "expr": "histogram_quantile(0.75, sum(rate(http_server_requests_seconds_bucket{application=\"$application\", instance=\"$instance\"}[1m])) by (le))",
              "format": "time_series",
              "intervalFactor": 1,
              "legendFormat": "75%",
              "refId": "C"
            },
            {
              "expr": "histogram_quantile(0.5, sum(rate(http_server_requests_seconds_bucket{application=\"$application\", instance=\"$instance\"}[1m])) by (le))",
              "format": "time_series",
              "intervalFactor": 1,
              "legendFormat": "50%",
              "refId": "D"
            }
    

    小结

    springboot的WebMvcMetricsAutoConfiguration给mvc提供了http.server.requests指标,类型为timer,当输出到prometheus的时候,默认是没有开启Histogram,输出的是一个_max的gauge,以及summary类型(_count及_sum)。

    相关文章

      网友评论

          本文标题:聊聊springboot的http.server.request

          本文链接:https://www.haomeiwen.com/subject/oecqgdtx.html