fix: prevent tracing internal requests.

The urllib3, library which we instrument, is used by the requests
package, and this last one is used by Instana Python Tracer to connect
with an Instana Agent.

When the INSTANA_ALLOW_EXIT_AS_ROOT environment variable is setup, the
internal requests are traced as EXIT_ROOT spans.

This commit prevents the internal requests to be traced by our lib.

Signed-off-by: Paulo Vital <paulo.vital@ibm.com>

src/instana/agent/host.py

@@ -40,6 +40,9 @@ class HostAgent(BaseAgent):
     """
     AGENT_DISCOVERY_PATH = "com.instana.plugin.python.discovery"
     AGENT_DATA_PATH = "com.instana.plugin.python.%d"
+    AGENT_HEADERS = {
+        "User-Agent": f"instana-python-tracer/{VERSION}",
+    }
 
     def __init__(self):
         super(HostAgent, self).__init__()
@@ -154,7 +157,8 @@ def is_agent_listening(self, host, port):
         result = False
         try:
             url = "http://%s:%s/" % (host, port)
-            response = self.client.get(url, timeout=5)
+            headers = self.AGENT_HEADERS
+            response = self.client.get(url, timeout=5, headers=headers)
 
             if 200 <= response.status_code < 300:
                 logger.debug("Instana host agent found on %s:%d", host, port)
@@ -174,9 +178,11 @@ def announce(self, discovery):
         """
         try:
             url = self.__discovery_url()
+            headers = {"Content-Type": "application/json"}
+            headers.update(self.AGENT_HEADERS)
             response = self.client.put(url,
                                        data=to_json(discovery),
-                                       headers={"Content-Type": "application/json"},
+                                       headers=headers,
                                        timeout=0.8)
         except Exception as exc:
             logger.debug("announce: connection error (%s)", type(exc))
@@ -224,10 +230,14 @@ def log_message_to_host_agent(self, message):
             payload["m"] = message
 
             url = self.__agent_logger_url()
+            headers = {
+                "Content-Type": "application/json",
+                "X-Log-Level": "INFO",
+            }
+            headers.update(self.AGENT_HEADERS)
             response = self.client.post(url,
                                        data=to_json(payload),
-                                       headers={"Content-Type": "application/json",
-                                                "X-Log-Level": "INFO"},
+                                       headers=headers,
                                        timeout=0.8)
 
             if 200 <= response.status_code <= 204:
@@ -241,7 +251,8 @@ def is_agent_ready(self):
         """
         ready = False
         try:
-            response = self.client.head(self.__data_url(), timeout=0.8)
+            headers = self.AGENT_HEADERS
+            response = self.client.head(self.__data_url(), timeout=0.8, headers=headers)
 
             if response.status_code == 200:
                 ready = True
@@ -259,9 +270,11 @@ def report_data_payload(self, payload):
             span_count = len(payload['spans'])
             if span_count > 0:
                 logger.debug("Reporting %d spans", span_count)
+                headers = {"Content-Type": "application/json"}
+                headers.update(self.AGENT_HEADERS)
                 response = self.client.post(self.__traces_url(),
                                             data=to_json(payload['spans']),
-                                            headers={"Content-Type": "application/json"},
+                                            headers=headers,
                                             timeout=0.8)
 
             if response is not None and 200 <= response.status_code <= 204:
@@ -271,19 +284,23 @@ def report_data_payload(self, payload):
             profile_count = len(payload['profiles'])
             if profile_count > 0:
                 logger.debug("Reporting %d profiles", profile_count)
+                headers = {"Content-Type": "application/json"}
+                headers.update(self.AGENT_HEADERS)
                 response = self.client.post(self.__profiles_url(),
                                             data=to_json(payload['profiles']),
-                                            headers={"Content-Type": "application/json"},
+                                            headers=headers,
                                             timeout=0.8)
 
             if response is not None and 200 <= response.status_code <= 204:
                 self.last_seen = datetime.now()
 
             # Report metrics
             metric_bundle = payload["metrics"]["plugins"][0]["data"]
+            headers = {"Content-Type": "application/json"}
+            headers.update(self.AGENT_HEADERS)
             response = self.client.post(self.__data_url(),
                                         data=to_json(metric_bundle),
-                                        headers={"Content-Type": "application/json"},
+                                        headers=headers,
                                         timeout=0.8)
 
             if response is not None and 200 <= response.status_code <= 204:
@@ -375,9 +392,11 @@ def __task_response(self, message_id, data):
 
             logger.debug("Task response is %s: %s", self.__response_url(message_id), payload)
 
+            headers = {"Content-Type": "application/json"}
+            headers.update(self.AGENT_HEADERS)
             response = self.client.post(self.__response_url(message_id),
                                         data=payload,
-                                        headers={"Content-Type": "application/json"},
+                                        headers=headers,
                                         timeout=0.8)
         except Exception as exc:
             logger.debug("__task_response: Instana host agent connection error (%s)", type(exc))

src/instana/instrumentation/urllib3.py

@@ -8,7 +8,7 @@
 
 from ..log import logger
 from ..singletons import agent
-from ..util.traceutils import get_tracer_tuple, tracing_is_off
+from ..util.traceutils import get_tracer_tuple, is_internal_request, tracing_is_off
 from ..util.secrets import strip_secrets_from_query
 
 try:
@@ -78,7 +78,7 @@ def collect_response(scope, response):
     def urlopen_with_instana(wrapped, instance, args, kwargs):
         tracer, parent_span, operation_name = get_tracer_tuple()
         # If we're not tracing, just return; boto3 has it's own visibility
-        if (tracing_is_off() or (operation_name == 'boto3')):
+        if tracing_is_off() or (operation_name == 'boto3') or is_internal_request(kwargs.get("headers", {})):
             return wrapped(*args, **kwargs)
 
         with tracer.start_active_span("urllib3", child_of=parent_span) as scope:

src/instana/util/traceutils.py

@@ -1,6 +1,7 @@
 # (c) Copyright IBM Corp. 2021
 # (c) Copyright Instana Inc. 2021
 
+from typing import Any, Dict
 from ..singletons import agent, tracer, async_tracer, tornado_tracer
 from ..log import logger
 
@@ -43,3 +44,17 @@ def get_tracer_tuple():
 
 def tracing_is_off():
     return not (bool(get_active_tracer()) or agent.options.allow_exit_as_root)
+
+
+def is_internal_request(headers: Dict[str, Any]) -> bool:
+    """
+    Check for the "instana-python-tracer/{VERSION}" string in the "User-Agent" 
+    header for identify internal requests.
+    """
+    # make all keys from the dict lowercase.
+    headers = dict(map(lambda kv: (kv[0].lower(), kv[1]), headers.items()))
+
+    if "user-agent" in headers.keys():
+        if "instana-python-tracer" in headers["user-agent"]:
+            return True
+    return False