Merge pull request #3664 from matrix-org/rav/federation_metrics

more metrics for the federation and appservice senders

changelog.d/3664.feature

@@ -0,0 +1 @@
+Add some metrics for the appservice and federation event sending loops

synapse/federation/transaction_queue.py

@@ -26,6 +26,8 @@ from synapse.api.errors import FederationDeniedError, HttpResponseException
 from synapse.handlers.presence import format_user_presence_state, get_interested_remotes
 from synapse.metrics import (
     LaterGauge,
+    event_processing_loop_counter,
+    event_processing_loop_room_count,
     events_processed_counter,
     sent_edus_counter,
     sent_transactions_counter,
@@ -253,7 +255,13 @@ class TransactionQueue(object):
                     synapse.metrics.event_processing_last_ts.labels(
                         "federation_sender").set(ts)
 
-                events_processed_counter.inc(len(events))
+                    events_processed_counter.inc(len(events))
+
+                    event_processing_loop_room_count.labels(
+                        "federation_sender"
+                    ).inc(len(events_by_room))
+
+                event_processing_loop_counter.labels("federation_sender").inc()
 
                 synapse.metrics.event_processing_positions.labels(
                     "federation_sender").set(next_token)

synapse/handlers/appservice.py

@@ -23,6 +23,10 @@ from twisted.internet import defer
 
 import synapse
 from synapse.api.constants import EventTypes
+from synapse.metrics import (
+    event_processing_loop_counter,
+    event_processing_loop_room_count,
+)
 from synapse.metrics.background_process_metrics import run_as_background_process
 from synapse.util.logcontext import make_deferred_yieldable, run_in_background
 from synapse.util.metrics import Measure
@@ -136,6 +140,12 @@ class ApplicationServicesHandler(object):
 
                     events_processed_counter.inc(len(events))
 
+                    event_processing_loop_room_count.labels(
+                        "appservice_sender"
+                    ).inc(len(events_by_room))
+
+                    event_processing_loop_counter.labels("appservice_sender").inc()
+
                     synapse.metrics.event_processing_lag.labels(
                         "appservice_sender").set(now - ts)
                     synapse.metrics.event_processing_last_ts.labels(

synapse/metrics/__init__.py

@@ -174,6 +174,19 @@ sent_transactions_counter = Counter("synapse_federation_client_sent_transactions
 
 events_processed_counter = Counter("synapse_federation_client_events_processed", "")
 
+event_processing_loop_counter = Counter(
+    "synapse_event_processing_loop_count",
+    "Event processing loop iterations",
+    ["name"],
+)
+
+event_processing_loop_room_count = Counter(
+    "synapse_event_processing_loop_room_count",
+    "Rooms seen per event processing loop iteration",
+    ["name"],
+)
+
+
 # Used to track where various components have processed in the event stream,
 # e.g. federation sending, appservice sending, etc.
 event_processing_positions = Gauge("synapse_event_processing_positions", "", ["name"])