From 5172c8c403d94ea5f184abc8b3c37dbd19a849bc Mon Sep 17 00:00:00 2001
From: reivilibre <oliverw@matrix.org>
Date: Wed, 11 Jan 2023 13:21:53 +0000
Subject: [PATCH] Faster remote room joins (worker mode): do not populate
 external hosts-in-room cache when sending events as this requires blocking
 for full state. (#14749)

Signed-off-by: Olivier Wilkinson (reivilibre) <oliverw@matrix.org>
Co-authored-by: Sean Quah <seanq@matrix.org>
---
 changelog.d/14749.misc      |  1 +
 synapse/handlers/message.py | 21 ++++++++++++++++-----
 2 files changed, 17 insertions(+), 5 deletions(-)
 create mode 100644 changelog.d/14749.misc

diff --git a/changelog.d/14749.misc b/changelog.d/14749.misc
new file mode 100644
index 0000000000..ff81325225
--- /dev/null
+++ b/changelog.d/14749.misc
@@ -0,0 +1 @@
+Faster remote room joins (worker mode): do not populate external hosts-in-room cache when sending events as this requires blocking for full state.
\ No newline at end of file
diff --git a/synapse/handlers/message.py b/synapse/handlers/message.py
index 88fc51a4c9..3278a695ed 100644
--- a/synapse/handlers/message.py
+++ b/synapse/handlers/message.py
@@ -1531,12 +1531,23 @@ class EventCreationHandler:
         external federation senders don't have to recalculate it themselves.
         """
 
-        for event, _ in events_and_context:
-            if not self._external_cache.is_enabled():
-                return
+        if not self._external_cache.is_enabled():
+            return
 
-            # If external cache is enabled we should always have this.
-            assert self._external_cache_joined_hosts_updates is not None
+        # If external cache is enabled we should always have this.
+        assert self._external_cache_joined_hosts_updates is not None
+
+        for event, event_context in events_and_context:
+            if event_context.partial_state:
+                # To populate the cache for a partial-state event, we either have to
+                # block until full state, which the code below does, or change the
+                # meaning of cache values to be the list of hosts to which we plan to
+                # send events and calculate that instead.
+                #
+                # The federation senders don't use the external cache when sending
+                # events in partial-state rooms anyway, so let's not bother populating
+                # the cache.
+                continue
 
             # We actually store two mappings, event ID -> prev state group,
             # state group -> joined hosts, which is much more space efficient
-- 
GitLab