Merge pull request #662 from matrix-org/erikj/state_cache

Make StateHandler._state_cache only store event_ids.

synapse/state.py

@@ -28,6 +28,7 @@ from collections import namedtuple
 
 import logging
 import hashlib
+import os
 
 logger = logging.getLogger(__name__)
 
@@ -35,8 +36,11 @@ logger = logging.getLogger(__name__)
 KeyStateTuple = namedtuple("KeyStateTuple", ("context", "type", "state_key"))
 
 
-SIZE_OF_CACHE = 1000
+CACHE_SIZE_FACTOR = float(os.environ.get("SYNAPSE_CACHE_FACTOR", 0.1))
-EVICTION_TIMEOUT_SECONDS = 20
+
+
+SIZE_OF_CACHE = int(1000 * CACHE_SIZE_FACTOR)
+EVICTION_TIMEOUT_SECONDS = 60 * 60
 
 
 class _StateCacheEntry(object):
@@ -86,16 +90,8 @@ class StateHandler(object):
         """
         event_ids = yield self.store.get_latest_event_ids_in_room(room_id)
 
-        cache = None
+        res = yield self.resolve_state_groups(room_id, event_ids)
-        if self._state_cache is not None:
+        state = res[1]
-            cache = self._state_cache.get(frozenset(event_ids), None)
-
-        if cache:
-            cache.ts = self.clock.time_msec()
-            state = cache.state
-        else:
-            res = yield self.resolve_state_groups(room_id, event_ids)
-            state = res[1]
 
         if event_type:
             defer.returnValue(state.get((event_type, state_key)))
@@ -187,20 +183,6 @@ class StateHandler(object):
         """
         logger.debug("resolve_state_groups event_ids %s", event_ids)
 
-        if self._state_cache is not None:
-            cache = self._state_cache.get(frozenset(event_ids), None)
-            if cache and cache.state_group:
-                cache.ts = self.clock.time_msec()
-                prev_state = cache.state.get((event_type, state_key), None)
-                if prev_state:
-                    prev_state = prev_state.event_id
-                    prev_states = [prev_state]
-                else:
-                    prev_states = []
-                defer.returnValue(
-                    (cache.state_group, cache.state, prev_states)
-                )
-
         state_groups = yield self.store.get_state_groups(
             room_id, event_ids
         )
@@ -210,7 +192,7 @@ class StateHandler(object):
             state_groups.keys()
         )
 
-        group_names = set(state_groups.keys())
+        group_names = frozenset(state_groups.keys())
         if len(group_names) == 1:
             name, state_list = state_groups.items().pop()
             state = {
@@ -224,29 +206,38 @@ class StateHandler(object):
             else:
                 prev_states = []
 
-            if self._state_cache is not None:
-                cache = _StateCacheEntry(
-                    state=state,
-                    state_group=name,
-                    ts=self.clock.time_msec()
-                )
-
-                self._state_cache[frozenset(event_ids)] = cache
-
             defer.returnValue((name, state, prev_states))
 
+        if self._state_cache is not None:
+            cache = self._state_cache.get(group_names, None)
+            if cache and cache.state_group:
+                cache.ts = self.clock.time_msec()
+
+                event_dict = yield self.store.get_events(cache.state.values())
+                state = {(e.type, e.state_key): e for e in event_dict.values()}
+
+                prev_state = state.get((event_type, state_key), None)
+                if prev_state:
+                    prev_state = prev_state.event_id
+                    prev_states = [prev_state]
+                else:
+                    prev_states = []
+                defer.returnValue(
+                    (cache.state_group, state, prev_states)
+                )
+
         new_state, prev_states = self._resolve_events(
             state_groups.values(), event_type, state_key
         )
 
         if self._state_cache is not None:
             cache = _StateCacheEntry(
-                state=new_state,
+                state={key: event.event_id for key, event in new_state.items()},
                 state_group=None,
                 ts=self.clock.time_msec()
             )
 
-            self._state_cache[frozenset(event_ids)] = cache
+            self._state_cache[group_names] = cache
 
         defer.returnValue((None, new_state, prev_states))
 

synapse/storage/events.py

@@ -151,6 +151,31 @@ class EventsStore(SQLBaseStore):
 
         defer.returnValue(events[0] if events else None)
 
+    @defer.inlineCallbacks
+    def get_events(self, event_ids, check_redacted=True,
+                   get_prev_content=False, allow_rejected=False):
+        """Get events from the database
+
+        Args:
+            event_ids (list): The event_ids of the events to fetch
+            check_redacted (bool): If True, check if event has been redacted
+                and redact it.
+            get_prev_content (bool): If True and event is a state event,
+                include the previous states content in the unsigned field.
+            allow_rejected (bool): If True return rejected events.
+
+        Returns:
+            Deferred : Dict from event_id to event.
+        """
+        events = yield self._get_events(
+            event_ids,
+            check_redacted=check_redacted,
+            get_prev_content=get_prev_content,
+            allow_rejected=allow_rejected,
+        )
+
+        defer.returnValue({e.event_id: e for e in events})
+
     @log_function
     def _persist_event_txn(self, txn, event, context,
                            is_new_state=True, current_state=None):

synapse/util/caches/lrucache.py

@@ -29,6 +29,16 @@ def enumerate_leaves(node, depth):
                 yield m
 
 
+class _Node(object):
+    __slots__ = ["prev_node", "next_node", "key", "value"]
+
+    def __init__(self, prev_node, next_node, key, value):
+        self.prev_node = prev_node
+        self.next_node = next_node
+        self.key = key
+        self.value = value
+
+
 class LruCache(object):
     """
     Least-recently-used cache.
@@ -38,10 +48,9 @@ class LruCache(object):
     def __init__(self, max_size, keylen=1, cache_type=dict):
         cache = cache_type()
         self.cache = cache  # Used for introspection.
-        list_root = []
+        list_root = _Node(None, None, None, None)
-        list_root[:] = [list_root, list_root, None, None]
+        list_root.next_node = list_root
-
+        list_root.prev_node = list_root
-        PREV, NEXT, KEY, VALUE = 0, 1, 2, 3
 
         lock = threading.Lock()
 
@@ -55,36 +64,36 @@ class LruCache(object):
 
         def add_node(key, value):
             prev_node = list_root
-            next_node = prev_node[NEXT]
+            next_node = prev_node.next_node
-            node = [prev_node, next_node, key, value]
+            node = _Node(prev_node, next_node, key, value)
-            prev_node[NEXT] = node
+            prev_node.next_node = node
-            next_node[PREV] = node
+            next_node.prev_node = node
             cache[key] = node
 
         def move_node_to_front(node):
-            prev_node = node[PREV]
+            prev_node = node.prev_node
-            next_node = node[NEXT]
+            next_node = node.next_node
-            prev_node[NEXT] = next_node
+            prev_node.next_node = next_node
-            next_node[PREV] = prev_node
+            next_node.prev_node = prev_node
             prev_node = list_root
-            next_node = prev_node[NEXT]
+            next_node = prev_node.next_node
-            node[PREV] = prev_node
+            node.prev_node = prev_node
-            node[NEXT] = next_node
+            node.next_node = next_node
-            prev_node[NEXT] = node
+            prev_node.next_node = node
-            next_node[PREV] = node
+            next_node.prev_node = node
 
         def delete_node(node):
-            prev_node = node[PREV]
+            prev_node = node.prev_node
-            next_node = node[NEXT]
+            next_node = node.next_node
-            prev_node[NEXT] = next_node
+            prev_node.next_node = next_node
-            next_node[PREV] = prev_node
+            next_node.prev_node = prev_node
 
         @synchronized
         def cache_get(key, default=None):
             node = cache.get(key, None)
             if node is not None:
                 move_node_to_front(node)
-                return node[VALUE]
+                return node.value
             else:
                 return default
 
@@ -93,25 +102,25 @@ class LruCache(object):
             node = cache.get(key, None)
             if node is not None:
                 move_node_to_front(node)
-                node[VALUE] = value
+                node.value = value
             else:
                 add_node(key, value)
                 if len(cache) > max_size:
-                    todelete = list_root[PREV]
+                    todelete = list_root.prev_node
                     delete_node(todelete)
-                    cache.pop(todelete[KEY], None)
+                    cache.pop(todelete.key, None)
 
         @synchronized
         def cache_set_default(key, value):
             node = cache.get(key, None)
             if node is not None:
-                return node[VALUE]
+                return node.value
             else:
                 add_node(key, value)
                 if len(cache) > max_size:
-                    todelete = list_root[PREV]
+                    todelete = list_root.prev_node
                     delete_node(todelete)
-                    cache.pop(todelete[KEY], None)
+                    cache.pop(todelete.key, None)
                 return value
 
         @synchronized
@@ -119,8 +128,8 @@ class LruCache(object):
             node = cache.get(key, None)
             if node:
                 delete_node(node)
-                cache.pop(node[KEY], None)
+                cache.pop(node.key, None)
-                return node[VALUE]
+                return node.value
             else:
                 return default
 
@@ -137,8 +146,8 @@ class LruCache(object):
 
         @synchronized
         def cache_clear():
-            list_root[NEXT] = list_root
+            list_root.next_node = list_root
-            list_root[PREV] = list_root
+            list_root.prev_node = list_root
             cache.clear()
 
         @synchronized