Merge branch 'grafana:main' into helm-chart

nokia · Feb 23, 2024 · 9f56afe · 9f56afe
2 parents d26d692 + cc085a6
commit 9f56afe
Show file tree

Hide file tree

Showing 27 changed files with 73 additions and 37 deletions.
diff --git a/bpf/go_common.h b/bpf/go_common.h
@@ -155,7 +155,6 @@ static __always_inline void server_trace_parent(void *goroutine_addr, tp_info_t
 static __always_inline u8 client_trace_parent(void *goroutine_addr, tp_info_t *tp_i, void *req_header) {
     // Get traceparent from the Request.Header
     u8 found_trace_id = 0;
-    u8 trace_id_exists = 0;
 
     // May get overriden when decoding existing traceparent or finding a server span, but otherwise we set sample ON
     tp_i->flags = 1;
@@ -164,7 +163,6 @@ static __always_inline u8 client_trace_parent(void *goroutine_addr, tp_info_t *t
         void *traceparent_ptr = extract_traceparent_from_req_headers(req_header);
         if (traceparent_ptr != NULL) {
             unsigned char buf[TP_MAX_VAL_LENGTH];
-            trace_id_exists = 1;
             long res = bpf_probe_read(buf, sizeof(buf), traceparent_ptr);
             if (res < 0) {
                 bpf_dbg_printk("can't copy traceparent header");
@@ -194,7 +192,7 @@ static __always_inline u8 client_trace_parent(void *goroutine_addr, tp_info_t *t
         urand_bytes(tp_i->span_id, SPAN_ID_SIZE_BYTES);
     }
 
-    return trace_id_exists;
+    return found_trace_id;
 }
 
 

diff --git a/bpf/go_grpc.c b/bpf/go_grpc.c
@@ -236,6 +236,7 @@ int uprobe_server_handleStream_return(struct pt_regs *ctx) {
 done:
     bpf_map_delete_elem(&ongoing_grpc_server_requests, &goroutine_addr);
     bpf_map_delete_elem(&ongoing_grpc_request_status, &goroutine_addr);
+    bpf_map_delete_elem(&go_trace_map, &goroutine_addr);
 
     return 0;
 }

diff --git a/bpf/go_nethttp.c b/bpf/go_nethttp.c
@@ -211,6 +211,7 @@ static __always_inline int writeHeaderHelper(struct pt_regs *ctx, u64 req_offset
 
 done:
     bpf_map_delete_elem(&ongoing_http_server_requests, &goroutine_addr);
+    bpf_map_delete_elem(&go_trace_map, &goroutine_addr);
     return 0;
 }
 
@@ -278,15 +279,15 @@ static __always_inline void roundTripStartHelper(struct pt_regs *ctx) {
     bpf_map_update_elem(&ongoing_http_client_requests_data, &goroutine_addr, &trace, BPF_ANY);
 
 #ifndef NO_HEADER_PROPAGATION
-    if (!existing_tp) {
+    //if (!existing_tp) {
         void *headers_ptr = 0;
         bpf_probe_read(&headers_ptr, sizeof(headers_ptr), (void*)(req + req_header_ptr_pos));
         bpf_dbg_printk("goroutine_addr %lx, req ptr %llx, headers_ptr %llx", goroutine_addr, req, headers_ptr);
 
         if (headers_ptr) {
             bpf_map_update_elem(&header_req_map, &headers_ptr, &goroutine_addr, BPF_ANY);
         }
-    }
+    //}
 #endif
 }
 
@@ -542,7 +543,7 @@ int uprobe_http2FramerWriteHeaders_returns(struct pt_regs *ctx) {
             bpf_probe_read(&n, sizeof(n), (void *)(w_ptr + 40));
             bpf_probe_read(&cap, sizeof(cap), (void *)(w_ptr + 24));
 
-            bpf_dbg_printk("Found f_info, this is the place to write to w = %llx, buf=%llx, n=%d, size=%d", w_ptr, buf_arr, n, cap);
+            bpf_dbg_printk("Found f_info, this is the place to write to w = %llx, buf=%llx, n=%lld, size=%lld", w_ptr, buf_arr, n, cap);
             if (buf_arr && n < (cap - HTTP2_ENCODED_HEADER_LEN)) {
                 uint8_t tp_str[TP_MAX_VAL_LENGTH];
 

diff --git a/docs/sources/configure/options.md b/docs/sources/configure/options.md
@@ -757,6 +757,17 @@ It is disabled by default to avoid cardinality explosion.
 The `buckets` object allows overriding the bucket boundaries of diverse histograms. See
 [Overriding histogram buckets](#overriding-histogram-buckets) section for more details.
 
+| YAML                    | Environment variable                                       | Type     | Default                     |
+|-------------------------|------------------------------------------------------------|----------|-----------------------------|
+| `histogram_aggregation` | `OTEL_EXPORTER_OTLP_METRICS_DEFAULT_HISTOGRAM_AGGREGATION` | `string` | `explicit_bucket_histogram` |
+
+Specifies the default aggregation to use for histogram instruments.
+
+Accepted values are:
+
+* `explicit_bucket_histogram` (default): use [Explicit Bucket Histogram Aggregation](https://opentelemetry.io/docs/specs/otel/metrics/sdk/#explicit-bucket-histogram-aggregation).
+* `base2_exponential_bucket_histogram`: use [Base2 Exponential Bucket Histogram Aggregation](https://opentelemetry.io/docs/specs/otel/metrics/sdk/#base2-exponential-bucket-histogram-aggregation).
+
 ### Overriding histogram buckets
 
 For both OpenTelemetry and Prometheus metrics exporters, you can override the histogram bucket
@@ -800,14 +811,14 @@ conventions recommend a different set of bucket boundaries.
 
 ### Use native histograms and exponential histograms
 
-For Prometheus [native histograms](https://prometheus.io/docs/concepts/metric_types/#histogram) are enabled if you have the `native-histograms` feature enabled.
-
-For OpenTelemetry you can use [exponential histograms](https://opentelemetry.io/docs/specs/otel/metrics/data-model/#exponentialhistogram) for the predefined histograms instead of defining the buckets manually.
-
-| YAML       | Environment variable                                                      | Type | Default |
-| ---------- | ------------------------------------------------------------------------- | ---- | ------- |
-| `use_exponential_histograms` | `BEYLA_OTEL_USE_EXPONENTIAL_HISTOGRAMS`  | `boolean` | `false` |
+For Prometheus, [native histograms](https://prometheus.io/docs/concepts/metric_types/#histogram) are enabled if you
+[enable the `native-histograms` feature in your Prometheus collector](https://prometheus.io/docs/prometheus/latest/feature_flags/#native-histograms).
 
+For OpenTelemetry you can use [exponential histograms](https://opentelemetry.io/docs/specs/otel/metrics/data-model/#exponentialhistogram)
+for the predefined histograms instead of defining the buckets manually. You need to set up the standard
+[OTEL_EXPORTER_OTLP_METRICS_DEFAULT_HISTOGRAM_AGGREGATION](https://opentelemetry.io/docs/specs/otel/metrics/sdk_exporters/otlp/#additional-configuration)
+environment variable. See the `histogram_aggregation` section in the [OTEL metrics exporter](#otel-metrics-exporter) section
+for more information.
 
 ## OTEL traces exporter
 

diff --git a/pkg/beyla/config.go b/pkg/beyla/config.go
@@ -43,11 +43,12 @@ var defaultConfig = Config{
 		},
 	},
 	Metrics: otel.MetricsConfig{
-		Protocol:          otel.ProtocolUnset,
-		MetricsProtocol:   otel.ProtocolUnset,
-		Interval:          5 * time.Second,
-		Buckets:           otel.DefaultBuckets,
-		ReportersCacheLen: ReporterLRUSize,
+		Protocol:             otel.ProtocolUnset,
+		MetricsProtocol:      otel.ProtocolUnset,
+		Interval:             5 * time.Second,
+		Buckets:              otel.DefaultBuckets,
+		ReportersCacheLen:    ReporterLRUSize,
+		HistogramAggregation: otel.AggregationExplicit,
 	},
 	Traces: otel.TracesConfig{
 		Protocol:           otel.ProtocolUnset,

diff --git a/pkg/beyla/config_test.go b/pkg/beyla/config_test.go
@@ -29,6 +29,7 @@ otel_metrics_export:
   endpoint: localhost:3030
   buckets:
     duration_histogram: [0, 1, 2]
+  histogram_aggregation: base2_exponential_bucket_histogram
 prometheus_export:
   buckets:
     request_size_histogram: [0, 10, 20, 22]
@@ -104,6 +105,7 @@ network:
 				DurationHistogram:    []float64{0, 1, 2},
 				RequestSizeHistogram: otel.DefaultBuckets.RequestSizeHistogram,
 			},
+			HistogramAggregation: "base2_exponential_bucket_histogram",
 		},
 		Traces: otel.TracesConfig{
 			Protocol:           otel.ProtocolUnset,

diff --git a/pkg/internal/ebpf/grpc/bpf_bpfel_arm64.o b/pkg/internal/ebpf/grpc/bpf_bpfel_arm64.o
diff --git a/pkg/internal/ebpf/grpc/bpf_bpfel_x86.o b/pkg/internal/ebpf/grpc/bpf_bpfel_x86.o
diff --git a/pkg/internal/ebpf/grpc/bpf_debug_bpfel_arm64.o b/pkg/internal/ebpf/grpc/bpf_debug_bpfel_arm64.o
diff --git a/pkg/internal/ebpf/grpc/bpf_debug_bpfel_x86.o b/pkg/internal/ebpf/grpc/bpf_debug_bpfel_x86.o
diff --git a/pkg/internal/ebpf/grpc/bpf_tp_bpfel_arm64.o b/pkg/internal/ebpf/grpc/bpf_tp_bpfel_arm64.o
diff --git a/pkg/internal/ebpf/grpc/bpf_tp_bpfel_x86.o b/pkg/internal/ebpf/grpc/bpf_tp_bpfel_x86.o
diff --git a/pkg/internal/ebpf/grpc/bpf_tp_debug_bpfel_arm64.o b/pkg/internal/ebpf/grpc/bpf_tp_debug_bpfel_arm64.o
diff --git a/pkg/internal/ebpf/grpc/bpf_tp_debug_bpfel_x86.o b/pkg/internal/ebpf/grpc/bpf_tp_debug_bpfel_x86.o
diff --git a/pkg/internal/ebpf/nethttp/bpf_bpfel_arm64.o b/pkg/internal/ebpf/nethttp/bpf_bpfel_arm64.o
diff --git a/pkg/internal/ebpf/nethttp/bpf_bpfel_x86.o b/pkg/internal/ebpf/nethttp/bpf_bpfel_x86.o
diff --git a/pkg/internal/ebpf/nethttp/bpf_debug_bpfel_arm64.o b/pkg/internal/ebpf/nethttp/bpf_debug_bpfel_arm64.o
diff --git a/pkg/internal/ebpf/nethttp/bpf_debug_bpfel_x86.o b/pkg/internal/ebpf/nethttp/bpf_debug_bpfel_x86.o
diff --git a/pkg/internal/ebpf/nethttp/bpf_tp_bpfel_arm64.o b/pkg/internal/ebpf/nethttp/bpf_tp_bpfel_arm64.o
diff --git a/pkg/internal/ebpf/nethttp/bpf_tp_bpfel_x86.o b/pkg/internal/ebpf/nethttp/bpf_tp_bpfel_x86.o
diff --git a/pkg/internal/ebpf/nethttp/bpf_tp_debug_bpfel_arm64.o b/pkg/internal/ebpf/nethttp/bpf_tp_debug_bpfel_arm64.o
diff --git a/pkg/internal/ebpf/nethttp/bpf_tp_debug_bpfel_x86.o b/pkg/internal/ebpf/nethttp/bpf_tp_debug_bpfel_x86.o
diff --git a/pkg/internal/export/otel/metrics.go b/pkg/internal/export/otel/metrics.go
@@ -39,6 +39,9 @@ const (
 
 	UsualPortGRPC = "4317"
 	UsualPortHTTP = "4318"
+
+	AggregationExplicit    = "explicit_bucket_histogram"
+	AggregationExponential = "base2_exponential_bucket_histogram"
 )
 
 type MetricsConfig struct {
@@ -59,8 +62,8 @@ type MetricsConfig struct {
 	ReportTarget   bool `yaml:"report_target" env:"BEYLA_METRICS_REPORT_TARGET"`
 	ReportPeerInfo bool `yaml:"report_peer" env:"BEYLA_METRICS_REPORT_PEER"`
 
-	Buckets                  Buckets `yaml:"buckets"`
-	UseExponentialHistograms bool    `yaml:"use_exponential_histograms" env:"BEYLA_OTEL_USE_EXPONENTIAL_HISTOGRAMS"`
+	Buckets              Buckets `yaml:"buckets"`
+	HistogramAggregation string  `yaml:"histogram_aggregation" env:"OTEL_EXPORTER_OTLP_METRICS_DEFAULT_HISTOGRAM_AGGREGATION"`
 
 	ReportersCacheLen int `yaml:"reporters_cache_len" env:"BEYLA_METRICS_REPORT_CACHE_LEN"`
 
@@ -171,21 +174,23 @@ func newMetricsReporter(ctx context.Context, cfg *MetricsConfig, ctxInfo *global
 }
 
 func (mr *MetricsReporter) newMetricSet(service svc.ID) (*Metrics, error) {
-	mlog().Debug("creating new Metrics reporter", "service", service)
+	mlog := mlog().With("service", service)
+	mlog.Debug("creating new Metrics reporter")
+	useExponentialHistograms := isExponentialAggregation(mr.cfg, mlog)
 	resources := otelResource(service)
 	m := Metrics{
 		ctx: mr.ctx,
 		provider: metric.NewMeterProvider(
 			metric.WithResource(resources),
 			metric.WithReader(metric.NewPeriodicReader(mr.exporter,
 				metric.WithInterval(mr.cfg.Interval))),
-			metric.WithView(otelHistogramConfig(HTTPServerDuration, mr.cfg.Buckets.DurationHistogram, mr.cfg.UseExponentialHistograms)),
-			metric.WithView(otelHistogramConfig(HTTPClientDuration, mr.cfg.Buckets.DurationHistogram, mr.cfg.UseExponentialHistograms)),
-			metric.WithView(otelHistogramConfig(RPCServerDuration, mr.cfg.Buckets.DurationHistogram, mr.cfg.UseExponentialHistograms)),
-			metric.WithView(otelHistogramConfig(RPCClientDuration, mr.cfg.Buckets.DurationHistogram, mr.cfg.UseExponentialHistograms)),
-			metric.WithView(otelHistogramConfig(SQLClientDuration, mr.cfg.Buckets.DurationHistogram, mr.cfg.UseExponentialHistograms)),
-			metric.WithView(otelHistogramConfig(HTTPServerRequestSize, mr.cfg.Buckets.RequestSizeHistogram, mr.cfg.UseExponentialHistograms)),
-			metric.WithView(otelHistogramConfig(HTTPClientRequestSize, mr.cfg.Buckets.RequestSizeHistogram, mr.cfg.UseExponentialHistograms)),
+			metric.WithView(otelHistogramConfig(HTTPServerDuration, mr.cfg.Buckets.DurationHistogram, useExponentialHistograms)),
+			metric.WithView(otelHistogramConfig(HTTPClientDuration, mr.cfg.Buckets.DurationHistogram, useExponentialHistograms)),
+			metric.WithView(otelHistogramConfig(RPCServerDuration, mr.cfg.Buckets.DurationHistogram, useExponentialHistograms)),
+			metric.WithView(otelHistogramConfig(RPCClientDuration, mr.cfg.Buckets.DurationHistogram, useExponentialHistograms)),
+			metric.WithView(otelHistogramConfig(SQLClientDuration, mr.cfg.Buckets.DurationHistogram, useExponentialHistograms)),
+			metric.WithView(otelHistogramConfig(HTTPServerRequestSize, mr.cfg.Buckets.RequestSizeHistogram, useExponentialHistograms)),
+			metric.WithView(otelHistogramConfig(HTTPClientRequestSize, mr.cfg.Buckets.RequestSizeHistogram, useExponentialHistograms)),
 		),
 	}
 	// time units for HTTP and GRPC durations are in seconds, according to the OTEL specification:
@@ -224,6 +229,20 @@ func (mr *MetricsReporter) newMetricSet(service svc.ID) (*Metrics, error) {
 	return &m, nil
 }
 
+func isExponentialAggregation(mc *MetricsConfig, mlog *slog.Logger) bool {
+	switch mc.HistogramAggregation {
+	case AggregationExponential:
+		return true
+	case AggregationExplicit:
+	// do nothing
+	default:
+		mlog.Warn("invalid value for histogram aggregation. Accepted values are: "+
+			AggregationExponential+", "+AggregationExplicit+" (default). Using default",
+			"value", mc.HistogramAggregation)
+	}
+	return false
+}
+
 // TODO: restore as private
 func InstantiateMetricsExporter(ctx context.Context, cfg *MetricsConfig, log *slog.Logger) (metric.Exporter, error) {
 	var err error

diff --git a/pkg/internal/goexec/offsets.json b/pkg/internal/goexec/offsets.json
@@ -118,7 +118,7 @@
       "target": {
         "versions": {
           "oldest": "1.40.0",
-          "newest": "1.61.1"
+          "newest": "1.62.0"
         },
         "offsets": [
           {
@@ -132,7 +132,7 @@
       "s": {
         "versions": {
           "oldest": "1.40.0",
-          "newest": "1.61.1"
+          "newest": "1.62.0"
         },
         "offsets": [
           {
@@ -146,7 +146,7 @@
       "ctx": {
         "versions": {
           "oldest": "1.40.0",
-          "newest": "1.61.1"
+          "newest": "1.62.0"
         },
         "offsets": [
           {
@@ -158,7 +158,7 @@
       "method": {
         "versions": {
           "oldest": "1.40.0",
-          "newest": "1.61.1"
+          "newest": "1.62.0"
         },
         "offsets": [
           {
@@ -170,7 +170,7 @@
       "st": {
         "versions": {
           "oldest": "1.40.0",
-          "newest": "1.61.1"
+          "newest": "1.62.0"
         },
         "offsets": [
           {
@@ -184,7 +184,7 @@
       "nextID": {
         "versions": {
           "oldest": "1.40.0",
-          "newest": "1.61.1"
+          "newest": "1.62.0"
         },
         "offsets": [
           {
@@ -226,7 +226,7 @@
       "peer": {
         "versions": {
           "oldest": "1.60.0",
-          "newest": "1.61.1"
+          "newest": "1.62.0"
         },
         "offsets": [
           {
@@ -252,7 +252,7 @@
       "Addr": {
         "versions": {
           "oldest": "1.60.0",
-          "newest": "1.61.1"
+          "newest": "1.62.0"
         },
         "offsets": [
           {
@@ -264,7 +264,7 @@
       "LocalAddr": {
         "versions": {
           "oldest": "1.60.0",
-          "newest": "1.61.1"
+          "newest": "1.62.0"
         },
         "offsets": [
           {

diff --git a/pkg/internal/netolly/ebpf/net_bpfel_arm64.go b/pkg/internal/netolly/ebpf/net_bpfel_arm64.go
diff --git a/pkg/internal/netolly/ebpf/net_bpfel_x86.go b/pkg/internal/netolly/ebpf/net_bpfel_x86.go
diff --git a/test/integration/suites_test.go b/test/integration/suites_test.go
@@ -135,6 +135,7 @@ func TestSuite_OldestGoVersion(t *testing.T) {
 }
 
 func TestSuite_UnsupportedGoVersion(t *testing.T) {
+	t.Skip("seems flaky, we need to look into this")
 	compose, err := docker.ComposeSuite("docker-compose-1.16.yml", path.Join(pathOutput, "test-suite-unsupported-go.log"))
 	require.NoError(t, err)
 	require.NoError(t, compose.Up())