Merge pull request #5476 from matrix-org/erikj/histogram_extremities

Add metrics for length of new extremities persisted.

3 files changed, 37 insertions, 1 deletions

diff --git a/changelog.d/5476.misc b/changelog.d/5476.misc
new file mode 100644
index 0000000000..7955c14744
--- /dev/null
+++ b/changelog.d/5476.misc
@@ -0,0 +1 @@
+Add new metrics for number of forward extremities being persisted and number of state groups involved in resolution.
diff --git a/synapse/state/__init__.py b/synapse/state/__init__.py
index fc20d1eaee..1b454a56a1 100644
--- a/synapse/state/__init__.py
+++ b/synapse/state/__init__.py
@@ -21,6 +21,7 @@ from six import iteritems, itervalues
 
 import attr
 from frozendict import frozendict
+from prometheus_client import Histogram
 
 from twisted.internet import defer
 
@@ -37,6 +38,14 @@ from synapse.util.metrics import Measure
 logger = logging.getLogger(__name__)
 
 
+# Metrics for number of state groups involved in a resolution.
+state_groups_histogram = Histogram(
+    "synapse_state_number_state_groups_in_resolution",
+    "Number of state groups used when performing a state resolution",
+    buckets=(1, 2, 3, 5, 7, 10, 15, 20, 50, 100, 200, 500, "+Inf"),
+)
+
+
 KeyStateTuple = namedtuple("KeyStateTuple", ("context", "type", "state_key"))
 
 
@@ -476,6 +485,8 @@ class StateResolutionHandler(object):
                 "Resolving state for %s with %d groups", room_id, len(state_groups_ids)
             )
 
+            state_groups_histogram.observe(len(state_groups_ids))
+
             # start by assuming we won't have any conflicted state, and build up the new
             # state map by iterating through the state groups. If we discover a conflict,
             # we give up and instead use `resolve_events_with_store`.
diff --git a/synapse/storage/events.py b/synapse/storage/events.py
index de5965569c..fefba39ea1 100644
--- a/synapse/storage/events.py
+++ b/synapse/storage/events.py
@@ -24,7 +24,7 @@ from six import iteritems, text_type
 from six.moves import range
 
 from canonicaljson import json
-from prometheus_client import Counter
+from prometheus_client import Counter, Histogram
 
 from twisted.internet import defer
 
@@ -74,6 +74,21 @@ state_delta_reuse_delta_counter = Counter(
     "synapse_storage_events_state_delta_reuse_delta", ""
 )
 
+# The number of forward extremities for each new event.
+forward_extremities_counter = Histogram(
+    "synapse_storage_events_forward_extremities_persisted",
+    "Number of forward extremities for each new event",
+    buckets=(1, 2, 3, 5, 7, 10, 15, 20, 50, 100, 200, 500, "+Inf"),
+)
+
+# The number of stale forward extremities for each new event. Stale extremities
+# are those that were in the previous set of extremities as well as the new.
+stale_forward_extremities_counter = Histogram(
+    "synapse_storage_events_stale_forward_extremities_persisted",
+    "Number of unchanged forward extremities for each new event",
+    buckets=(0, 1, 2, 3, 5, 7, 10, 15, 20, 50, 100, 200, 500, "+Inf"),
+)
+
 
 def encode_json(json_object):
     """
@@ -541,6 +556,8 @@ class EventsStore(
             and not event.internal_metadata.is_soft_failed()
         ]
 
+        latest_event_ids = set(latest_event_ids)
+
         # start with the existing forward extremities
         result = set(latest_event_ids)
 
@@ -564,6 +581,13 @@ class EventsStore(
         )
         result.difference_update(existing_prevs)
 
+        # We only update metrics for events that change forward extremities
+        # (e.g. we ignore backfill/outliers/etc)
+        if result != latest_event_ids:
+            forward_extremities_counter.observe(len(result))
+            stale = latest_event_ids & result
+            stale_forward_extremities_counter.observe(len(stale))
+
         defer.returnValue(result)
 
     @defer.inlineCallbacks