Allow forcing full OTEL trace by setting query param

backend/src/baserow/core/telemetry/env_overrides_parser.py

@@ -8,15 +8,16 @@ from opentelemetry.sdk.trace.sampling import (
     DEFAULT_ON,
     ParentBasedTraceIdRatio,
     Sampler,
-    TraceIdRatioBased,
 )
 
+from .sampling import ForcableTraceIdRatioBased
+
 _KNOWN_SAMPLERS = {
     "always_on": ALWAYS_ON,
     "always_off": ALWAYS_OFF,
     "parentbased_always_on": DEFAULT_ON,
     "parentbased_always_off": DEFAULT_OFF,
-    "traceidratio": TraceIdRatioBased,
+    "traceidratio": ForcableTraceIdRatioBased,
     "parentbased_traceidratio": ParentBasedTraceIdRatio,
 }
 

backend/src/baserow/core/telemetry/sampling.py

@@ -0,0 +1,46 @@
+from typing import Optional
+
+from opentelemetry import trace
+from opentelemetry.context import Context
+from opentelemetry.sdk.trace import Span
+from opentelemetry.sdk.trace.sampling import _KNOWN_SAMPLERS as ROOT_KNOWN_SAMPLERS
+from opentelemetry.sdk.trace.sampling import (
+    Decision,
+    SamplingResult,
+    TraceIdRatioBased,
+    _get_parent_trace_state,
+)
+
+
+class ForcableTraceIdRatioBased(TraceIdRatioBased):
+    def should_sample(
+        self,
+        parent_context: Optional["Context"],
+        trace_id: int,
+        name: str,
+        kind=None,
+        attributes=None,
+        links=None,
+        trace_state=None,
+    ) -> SamplingResult:
+        current_span = trace.get_current_span(parent_context)
+
+        if isinstance(current_span, Span):
+            span_attributes = current_span.attributes
+            # Check if the HTTP request contains the query parameter to force full
+            # tracing, and if so do the full trace.
+            http_url = span_attributes.get("http.url", "")
+            if "force_full_otel_trace=true" in http_url:
+                return SamplingResult(
+                    Decision.RECORD_AND_SAMPLE,
+                    span_attributes,
+                    _get_parent_trace_state(parent_context),
+                )
+        # If the full trace query parameter is not provided, then fallback on the
+        # original TraceIdRatioBased class.
+        return super().should_sample(
+            parent_context, trace_id, name, kind, attributes, links, trace_state
+        )
+
+
+ROOT_KNOWN_SAMPLERS["traceidratio"] = ForcableTraceIdRatioBased

backend/tests/baserow/core/telemetry/test_env_overrides_parser.py

@@ -4,10 +4,10 @@ from opentelemetry.sdk.trace.sampling import (
     DEFAULT_OFF,
     DEFAULT_ON,
     ParentBasedTraceIdRatio,
-    TraceIdRatioBased,
 )
 
 from baserow.core.telemetry.env_overrides_parser import get_sampler_overrides_from_str
+from baserow.core.telemetry.sampling import ForcableTraceIdRatioBased
 
 
 def test_invalid_sampler_strings():
@@ -52,7 +52,7 @@ def test_valid_sampler_strings():
 
 def test_valid_sampler_with_args():
     for_trace_id = get_sampler_overrides_from_str("my.module=traceidratio@0.4")
-    assert type(for_trace_id["my.module"]) is TraceIdRatioBased
+    assert type(for_trace_id["my.module"]) is ForcableTraceIdRatioBased
     assert for_trace_id["my.module"].rate == 0.4
     for_parent_trace_id = get_sampler_overrides_from_str(
         "my.module=parentbased_traceidratio@0.4"

deploy/otel/otel-collector-config.yaml

@@ -2,7 +2,8 @@ receivers:
   otlp:
     protocols:
       grpc: # port 4317
-      http: # port 4318
+      http:
+        endpoint: "0.0.0.0:4318"
 
 processors:
   # Disable sending spans for the sync templates task as it contains a huge number
@@ -55,4 +56,4 @@ service:
       exporters: [otlp,logging]
 
 extensions:
-  health_check:
+  health_check: