Add extra spans to remote join path

2 files changed, 32 insertions, 24 deletions

diff --git a/synapse/federation/federation_client.py b/synapse/federation/federation_client.py
index b7a10da15a..a837c18726 100644
--- a/synapse/federation/federation_client.py
+++ b/synapse/federation/federation_client.py
@@ -55,6 +55,7 @@ from synapse.api.room_versions import (
 from synapse.events import EventBase, builder
 from synapse.federation.federation_base import FederationBase, event_from_pdu_json
 from synapse.federation.transport.client import SendJoinResponse
+from synapse.logging.opentracing import start_active_span
 from synapse.logging.utils import log_function
 from synapse.types import JsonDict, get_domain_from_id
 from synapse.util.async_helpers import concurrently_execute
@@ -732,7 +733,8 @@ class FederationClient(FederationBase):
         """
 
         async def send_request(destination) -> SendJoinResult:
-            response = await self._do_send_join(room_version, destination, pdu)
+            with start_active_span("_do_send_join"):
+                response = await self._do_send_join(room_version, destination, pdu)
 
             # If an event was returned (and expected to be returned):
             #
@@ -792,19 +794,21 @@ class FederationClient(FederationBase):
             valid_pdus_map: Dict[str, EventBase] = {}
 
             async def _execute(pdu: EventBase) -> None:
-                valid_pdu = await self._check_sigs_and_hash_and_fetch_one(
-                    pdu=pdu,
-                    origin=destination,
-                    outlier=True,
-                    room_version=room_version,
-                )
+                with start_active_span("_check_sigs_and_hash_and_fetch_one"):
+                    valid_pdu = await self._check_sigs_and_hash_and_fetch_one(
+                        pdu=pdu,
+                        origin=destination,
+                        outlier=True,
+                        room_version=room_version,
+                    )
 
                 if valid_pdu:
                     valid_pdus_map[valid_pdu.event_id] = valid_pdu
 
-            await concurrently_execute(
-                _execute, itertools.chain(state, auth_chain), 10000
-            )
+            with start_active_span("check_sigs"):
+                await concurrently_execute(
+                    _execute, itertools.chain(state, auth_chain), 10000
+                )
 
             # NB: We *need* to copy to ensure that we don't have multiple
             # references being passed on, as that causes... issues.
diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py
index 8197b60b76..2e94852a72 100644
--- a/synapse/handlers/federation.py
+++ b/synapse/handlers/federation.py
@@ -72,6 +72,7 @@ from synapse.logging.context import (
     preserve_fn,
     run_in_background,
 )
+from synapse.logging.opentracing import start_active_span
 from synapse.logging.utils import log_function
 from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.replication.http.devices import ReplicationUserDevicesResyncRestServlet
@@ -1462,14 +1463,15 @@ class FederationHandler(BaseHandler):
 
         logger.debug("Joining %s to %s", joinee, room_id)
 
-        origin, event, room_version_obj = await self._make_and_verify_event(
-            target_hosts,
-            room_id,
-            joinee,
-            "join",
-            content,
-            params={"ver": KNOWN_ROOM_VERSIONS},
-        )
+        with start_active_span("make_join"):
+            origin, event, room_version_obj = await self._make_and_verify_event(
+                target_hosts,
+                room_id,
+                joinee,
+                "join",
+                content,
+                params={"ver": KNOWN_ROOM_VERSIONS},
+            )
 
         # This shouldn't happen, because the RoomMemberHandler has a
         # linearizer lock which only allows one operation per user per room
@@ -1490,9 +1492,10 @@ class FederationHandler(BaseHandler):
             except ValueError:
                 pass
 
-            ret = await self.federation_client.send_join(
-                host_list, event, room_version_obj
-            )
+            with start_active_span("send_join"):
+                ret = await self.federation_client.send_join(
+                    host_list, event, room_version_obj
+                )
 
             event = ret.event
             origin = ret.origin
@@ -1519,9 +1522,10 @@ class FederationHandler(BaseHandler):
                 room_version=room_version_obj,
             )
 
-            max_stream_id = await self._persist_auth_tree(
-                origin, room_id, auth_chain, state, event, room_version_obj
-            )
+            with start_active_span("_persist_auth_tree"):
+                max_stream_id = await self._persist_auth_tree(
+                    origin, room_id, auth_chain, state, event, room_version_obj
+                )
 
             # We wait here until this instance has seen the events come down
             # replication (if we're using replication) as the below uses caches.