From dbce28b2f14e5a4d0db9f090caf390ea998d1e59 Mon Sep 17 00:00:00 2001
From: Erik Johnston <erik@matrix.org>
Date: Thu, 30 Jun 2022 15:08:40 +0100
Subject: [PATCH] Fix unread counts on large servers (#13140)

---
 changelog.d/13140.bugfix                      |  1 +
 .../databases/main/event_push_actions.py      | 48 ++++++++++---------
 tests/storage/test_event_push_actions.py      | 12 ++---
 3 files changed, 32 insertions(+), 29 deletions(-)
 create mode 100644 changelog.d/13140.bugfix

diff --git a/changelog.d/13140.bugfix b/changelog.d/13140.bugfix
new file mode 100644
index 000000000..cb0586e39
--- /dev/null
+++ b/changelog.d/13140.bugfix
@@ -0,0 +1 @@
+Fix unread counts for users on large servers. Introduced in v1.62.0rc1.
diff --git a/synapse/storage/databases/main/event_push_actions.py b/synapse/storage/databases/main/event_push_actions.py
index 3fadd0930..7d4754b3d 100644
--- a/synapse/storage/databases/main/event_push_actions.py
+++ b/synapse/storage/databases/main/event_push_actions.py
@@ -854,18 +854,20 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, EventsWorkerStore, SQLBas
 
         limit = 100
 
-        min_stream_id = self.db_pool.simple_select_one_onecol_txn(
+        min_receipts_stream_id = self.db_pool.simple_select_one_onecol_txn(
             txn,
             table="event_push_summary_last_receipt_stream_id",
             keyvalues={},
             retcol="stream_id",
         )
 
+        max_receipts_stream_id = self._receipts_id_gen.get_current_token()
+
         sql = """
             SELECT r.stream_id, r.room_id, r.user_id, e.stream_ordering
             FROM receipts_linearized AS r
             INNER JOIN events AS e USING (event_id)
-            WHERE r.stream_id > ? AND user_id LIKE ?
+            WHERE ? < r.stream_id AND r.stream_id <= ? AND user_id LIKE ?
             ORDER BY r.stream_id ASC
             LIMIT ?
         """
@@ -877,13 +879,21 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, EventsWorkerStore, SQLBas
         txn.execute(
             sql,
             (
-                min_stream_id,
+                min_receipts_stream_id,
+                max_receipts_stream_id,
                 user_filter,
                 limit,
             ),
         )
         rows = txn.fetchall()
 
+        old_rotate_stream_ordering = self.db_pool.simple_select_one_onecol_txn(
+            txn,
+            table="event_push_summary_stream_ordering",
+            keyvalues={},
+            retcol="stream_ordering",
+        )
+
         # For each new read receipt we delete push actions from before it and
         # recalculate the summary.
         for _, room_id, user_id, stream_ordering in rows:
@@ -902,13 +912,6 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, EventsWorkerStore, SQLBas
                 (room_id, user_id, stream_ordering),
             )
 
-            old_rotate_stream_ordering = self.db_pool.simple_select_one_onecol_txn(
-                txn,
-                table="event_push_summary_stream_ordering",
-                keyvalues={},
-                retcol="stream_ordering",
-            )
-
             notif_count, unread_count = self._get_notif_unread_count_for_user_room(
                 txn, room_id, user_id, stream_ordering, old_rotate_stream_ordering
             )
@@ -927,18 +930,19 @@ class EventPushActionsWorkerStore(ReceiptsWorkerStore, EventsWorkerStore, SQLBas
 
         # We always update `event_push_summary_last_receipt_stream_id` to
         # ensure that we don't rescan the same receipts for remote users.
-        #
-        # This requires repeatable read to be safe, as we need the
-        # `MAX(stream_id)` to not include any new rows that have been committed
-        # since the start of the transaction (since those rows won't have been
-        # returned by the query above). Alternatively we could query the max
-        # stream ID at the start of the transaction and bound everything by
-        # that.
-        txn.execute(
-            """
-            UPDATE event_push_summary_last_receipt_stream_id
-            SET stream_id = (SELECT COALESCE(MAX(stream_id), 0) FROM receipts_linearized)
-            """
+
+        upper_limit = max_receipts_stream_id
+        if len(rows) >= limit:
+            # If we pulled out a limited number of rows we only update the
+            # position to the last receipt we processed, so we continue
+            # processing the rest next iteration.
+            upper_limit = rows[-1][0]
+
+        self.db_pool.simple_update_txn(
+            txn,
+            table="event_push_summary_last_receipt_stream_id",
+            keyvalues={},
+            updatevalues={"stream_id": upper_limit},
         )
 
         return len(rows) < limit
diff --git a/tests/storage/test_event_push_actions.py b/tests/storage/test_event_push_actions.py
index ef069a811..684485ae0 100644
--- a/tests/storage/test_event_push_actions.py
+++ b/tests/storage/test_event_push_actions.py
@@ -134,15 +134,12 @@ class EventPushActionsStoreTestCase(HomeserverTestCase):
             last_read_stream_ordering[0] = stream
 
             self.get_success(
-                self.store.db_pool.runInteraction(
-                    "",
-                    self.store._insert_linearized_receipt_txn,
+                self.store.insert_receipt(
                     room_id,
                     "m.read",
-                    user_id,
-                    f"$test{stream}:example.com",
-                    {},
-                    stream,
+                    user_id=user_id,
+                    event_ids=[f"$test{stream}:example.com"],
+                    data={},
                 )
             )
 
@@ -166,6 +163,7 @@ class EventPushActionsStoreTestCase(HomeserverTestCase):
 
         _inject_actions(6, PlAIN_NOTIF)
         _rotate(7)
+        _assert_counts(1, 0)
 
         self.get_success(
             self.store.db_pool.simple_delete(