Merge pull request #4006 from matrix-org/erikj/purge_state_groups

Delete unreferenced state groups during purge

changelog.d/4006.misc

@@ -0,0 +1 @@
+Delete unreferenced state groups during history purge

synapse/storage/events.py

@@ -38,6 +38,7 @@ from synapse.state import StateResolutionStore
 from synapse.storage.background_updates import BackgroundUpdateStore
 from synapse.storage.event_federation import EventFederationStore
 from synapse.storage.events_worker import EventsWorkerStore
+from synapse.storage.state import StateGroupWorkerStore
 from synapse.types import RoomStreamToken, get_domain_from_id
 from synapse.util import batch_iter
 from synapse.util.async_helpers import ObservableDeferred
@@ -205,7 +206,8 @@ def _retry_on_integrity_error(func):
 
 # inherits from EventFederationStore so that we can call _update_backward_extremities
 # and _handle_mult_prev_events (though arguably those could both be moved in here)
-class EventsStore(EventFederationStore, EventsWorkerStore, BackgroundUpdateStore):
+class EventsStore(StateGroupWorkerStore, EventFederationStore, EventsWorkerStore,
+                  BackgroundUpdateStore):
     EVENT_ORIGIN_SERVER_TS_NAME = "event_origin_server_ts"
     EVENT_FIELDS_SENDER_URL_UPDATE_NAME = "event_fields_sender_url"
 
@@ -2034,54 +2036,36 @@ class EventsStore(EventFederationStore, EventsWorkerStore, BackgroundUpdateStore
 
         logger.info("[purge] finding redundant state groups")
 
-        # Get all state groups that are only referenced by events that are
+        # Get all state groups that are referenced by events that are to be
-        # to be deleted.
+        # deleted. We then go and check if they are referenced by other events
-        # This works by first getting state groups that we may want to delete,
+        # or state groups, and if not we delete them.
-        # joining against event_to_state_groups to get events that use that
-        # state group, then left joining against events_to_purge again. Any
-        # state group where the left join produce *no nulls* are referenced
-        # only by events that are going to be purged.
         txn.execute("""
-            SELECT state_group FROM
-            (
             SELECT DISTINCT state_group FROM events_to_purge
             INNER JOIN event_to_state_groups USING (event_id)
-            ) AS sp
-            INNER JOIN event_to_state_groups USING (state_group)
-            LEFT JOIN events_to_purge AS ep USING (event_id)
-            GROUP BY state_group
-            HAVING SUM(CASE WHEN ep.event_id IS NULL THEN 1 ELSE 0 END) = 0
         """)
 
-        state_rows = txn.fetchall()
+        referenced_state_groups = set(sg for sg, in txn)
-        logger.info("[purge] found %i redundant state groups", len(state_rows))
+        logger.info(
-
+            "[purge] found %i referenced state groups",
-        # make a set of the redundant state groups, so that we can look them up
+            len(referenced_state_groups),
-        # efficiently
-        state_groups_to_delete = set([sg for sg, in state_rows])
-
-        # Now we get all the state groups that rely on these state groups
-        logger.info("[purge] finding state groups which depend on redundant"
-                    " state groups")
-        remaining_state_groups = []
-        for i in range(0, len(state_rows), 100):
-            chunk = [sg for sg, in state_rows[i:i + 100]]
-            # look for state groups whose prev_state_group is one we are about
-            # to delete
-            rows = self._simple_select_many_txn(
-                txn,
-                table="state_group_edges",
-                column="prev_state_group",
-                iterable=chunk,
-                retcols=["state_group"],
-                keyvalues={},
         )
-            remaining_state_groups.extend(
-                row["state_group"] for row in rows
 
-                # exclude state groups we are about to delete: no point in
+        logger.info("[purge] finding state groups that can be deleted")
-                # updating them
+
-                if row["state_group"] not in state_groups_to_delete
+        state_groups_to_delete, remaining_state_groups = (
+            self._find_unreferenced_groups_during_purge(
+                txn, referenced_state_groups,
+            )
+        )
+
+        logger.info(
+            "[purge] found %i state groups to delete",
+            len(state_groups_to_delete),
+        )
+
+        logger.info(
+            "[purge] de-delta-ing %i remaining state groups",
+            len(remaining_state_groups),
         )
 
         # Now we turn the state groups that reference to-be-deleted state
@@ -2127,11 +2111,11 @@ class EventsStore(EventFederationStore, EventsWorkerStore, BackgroundUpdateStore
         logger.info("[purge] removing redundant state groups")
         txn.executemany(
             "DELETE FROM state_groups_state WHERE state_group = ?",
-            state_rows
+            ((sg,) for sg in state_groups_to_delete),
         )
         txn.executemany(
             "DELETE FROM state_groups WHERE id = ?",
-            state_rows
+            ((sg,) for sg in state_groups_to_delete),
         )
 
         logger.info("[purge] removing events from event_to_state_groups")
@@ -2227,6 +2211,85 @@ class EventsStore(EventFederationStore, EventsWorkerStore, BackgroundUpdateStore
 
         logger.info("[purge] done")
 
+    def _find_unreferenced_groups_during_purge(self, txn, state_groups):
+        """Used when purging history to figure out which state groups can be
+        deleted and which need to be de-delta'ed (due to one of its prev groups
+        being scheduled for deletion).
+
+        Args:
+            txn
+            state_groups (set[int]): Set of state groups referenced by events
+                that are going to be deleted.
+
+        Returns:
+            tuple[set[int], set[int]]: The set of state groups that can be
+            deleted and the set of state groups that need to be de-delta'ed
+        """
+        # Graph of state group -> previous group
+        graph = {}
+
+        # Set of events that we have found to be referenced by events
+        referenced_groups = set()
+
+        # Set of state groups we've already seen
+        state_groups_seen = set(state_groups)
+
+        # Set of state groups to handle next.
+        next_to_search = set(state_groups)
+        while next_to_search:
+            # We bound size of groups we're looking up at once, to stop the
+            # SQL query getting too big
+            if len(next_to_search) < 100:
+                current_search = next_to_search
+                next_to_search = set()
+            else:
+                current_search = set(itertools.islice(next_to_search, 100))
+                next_to_search -= current_search
+
+            # Check if state groups are referenced
+            sql = """
+                SELECT DISTINCT state_group FROM event_to_state_groups
+                LEFT JOIN events_to_purge AS ep USING (event_id)
+                WHERE state_group IN (%s) AND ep.event_id IS NULL
+            """ % (",".join("?" for _ in current_search),)
+            txn.execute(sql, list(current_search))
+
+            referenced = set(sg for sg, in txn)
+            referenced_groups |= referenced
+
+            # We don't continue iterating up the state group graphs for state
+            # groups that are referenced.
+            current_search -= referenced
+
+            rows = self._simple_select_many_txn(
+                txn,
+                table="state_group_edges",
+                column="prev_state_group",
+                iterable=current_search,
+                keyvalues={},
+                retcols=("prev_state_group", "state_group",),
+            )
+
+            prevs = set(row["state_group"] for row in rows)
+            # We don't bother re-handling groups we've already seen
+            prevs -= state_groups_seen
+            next_to_search |= prevs
+            state_groups_seen |= prevs
+
+            for row in rows:
+                # Note: Each state group can have at most one prev group
+                graph[row["state_group"]] = row["prev_state_group"]
+
+        to_delete = state_groups_seen - referenced_groups
+
+        to_dedelta = set()
+        for sg in referenced_groups:
+            prev_sg = graph.get(sg)
+            if prev_sg and prev_sg in to_delete:
+                to_dedelta.add(sg)
+
+        return to_delete, to_dedelta
+
     @defer.inlineCallbacks
     def is_event_after(self, event_id1, event_id2):
         """Returns True if event_id1 is after event_id2 in the stream

synapse/storage/prepare_database.py

@@ -25,7 +25,7 @@ logger = logging.getLogger(__name__)
 
 # Remember to update this number every time a change is made to database
 # schema files, so the users will be informed on server restarts.
-SCHEMA_VERSION = 51
+SCHEMA_VERSION = 52
 
 dir_path = os.path.abspath(os.path.dirname(__file__))
 

synapse/storage/schema/delta/52/add_event_to_state_group_index.sql

@@ -0,0 +1,19 @@
+/* Copyright 2018 New Vector Ltd
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *    http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+-- This is needed to efficiently check for unreferenced state groups during
+-- purge. Added events_to_state_group(state_group) index
+INSERT into background_updates (update_name, progress_json)
+    VALUES ('event_to_state_groups_sg_index', '{}');

synapse/storage/state.py

@@ -1257,6 +1257,7 @@ class StateStore(StateGroupWorkerStore, BackgroundUpdateStore):
     STATE_GROUP_DEDUPLICATION_UPDATE_NAME = "state_group_state_deduplication"
     STATE_GROUP_INDEX_UPDATE_NAME = "state_group_state_type_index"
     CURRENT_STATE_INDEX_UPDATE_NAME = "current_state_members_idx"
+    EVENT_STATE_GROUP_INDEX_UPDATE_NAME = "event_to_state_groups_sg_index"
 
     def __init__(self, db_conn, hs):
         super(StateStore, self).__init__(db_conn, hs)
@@ -1275,6 +1276,12 @@ class StateStore(StateGroupWorkerStore, BackgroundUpdateStore):
             columns=["state_key"],
             where_clause="type='m.room.member'",
         )
+        self.register_background_index_update(
+            self.EVENT_STATE_GROUP_INDEX_UPDATE_NAME,
+            index_name="event_to_state_groups_sg_index",
+            table="event_to_state_groups",
+            columns=["state_group"],
+        )
 
     def _store_event_state_mappings_txn(self, txn, events_and_contexts):
         state_groups = {}