From 3f49e131d9a31d5bf198f2aa6b102562b2e747d0 Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Tue, 27 Mar 2018 10:30:03 +0100 Subject: Add counter metrics for calculating state delta This will allow us to measure how often we calculate state deltas in event persistence that we would have been able to calculate at the same time we calculated the state for the event. --- synapse/storage/events.py | 31 ++++++++++++++++++++++++++++++- 1 file changed, 30 insertions(+), 1 deletion(-) (limited to 'synapse/storage/events.py') diff --git a/synapse/storage/events.py b/synapse/storage/events.py index 85ce6bea1a..f3a362db20 100644 --- a/synapse/storage/events.py +++ b/synapse/storage/events.py @@ -52,6 +52,15 @@ persist_event_counter = metrics.register_counter("persisted_events") event_counter = metrics.register_counter( "persisted_events_sep", labels=["type", "origin_type", "origin_entity"] ) +state_delta_counter = metrics.register_counter( + "state_delta", +) +state_delta_single_event_counter = metrics.register_counter( + "state_delta_single_event", +) +state_delta_reuse_delta_counter = metrics.register_counter( + "state_delta_reuse_delta", +) def encode_json(json_object): @@ -368,7 +377,8 @@ class EventsStore(EventsWorkerStore): room_id, ev_ctx_rm, latest_event_ids ) - if new_latest_event_ids == set(latest_event_ids): + latest_event_ids = set(latest_event_ids) + if new_latest_event_ids == latest_event_ids: # No change in extremities, so no change in state continue @@ -389,6 +399,25 @@ class EventsStore(EventsWorkerStore): if all_single_prev_not_state: continue + state_delta_counter.inc() + if len(new_latest_event_ids) == 1: + state_delta_single_event_counter.inc() + + # This is a fairly handwavey check to see if we could + # have guessed what the delta would have been when + # processing one of these events. + # What we're interested in is if the latest extremities + # were the same when we created the event as they are + # now. We guess this by looking at the prev events and + # checking if they match up, as when this server creates + # a new event it will use the extremities as the prev + # events. + for ev, _ in ev_ctx_rm: + prev_event_ids = set(e for e, _ in ev.prev_events) + if latest_event_ids == prev_event_ids: + state_delta_reuse_delta_counter.inc() + break + logger.info( "Calculating state delta for room %s", room_id, ) -- cgit 1.5.1 From e70287cff3adeeeff2a7a1e95fc5845feb365710 Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Tue, 27 Mar 2018 13:13:38 +0100 Subject: Comment --- synapse/storage/events.py | 7 +++++++ 1 file changed, 7 insertions(+) (limited to 'synapse/storage/events.py') diff --git a/synapse/storage/events.py b/synapse/storage/events.py index f3a362db20..ee8b3c46d3 100644 --- a/synapse/storage/events.py +++ b/synapse/storage/events.py @@ -52,12 +52,19 @@ persist_event_counter = metrics.register_counter("persisted_events") event_counter = metrics.register_counter( "persisted_events_sep", labels=["type", "origin_type", "origin_entity"] ) + +# The number of times we are recalculating the current state state_delta_counter = metrics.register_counter( "state_delta", ) +# The number of times we are recalculating state when there is only a +# single forward extremity state_delta_single_event_counter = metrics.register_counter( "state_delta_single_event", ) +# The number of times we are reculating state when we could have resonably +# calculated the delta when we calculated the state for an event we were +# persisting. state_delta_reuse_delta_counter = metrics.register_counter( "state_delta_reuse_delta", ) -- cgit 1.5.1 From 152c2ac19e6af373a9e79a8f4da80d9e77a5396f Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Tue, 27 Mar 2018 13:13:46 +0100 Subject: Fix indent --- synapse/storage/events.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'synapse/storage/events.py') diff --git a/synapse/storage/events.py b/synapse/storage/events.py index ee8b3c46d3..95b36d669b 100644 --- a/synapse/storage/events.py +++ b/synapse/storage/events.py @@ -423,7 +423,7 @@ class EventsStore(EventsWorkerStore): prev_event_ids = set(e for e, _ in ev.prev_events) if latest_event_ids == prev_event_ids: state_delta_reuse_delta_counter.inc() - break + break logger.info( "Calculating state delta for room %s", room_id, -- cgit 1.5.1 From 800cfd5774d9726a00c53f2d84ce0286ab64c17b Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Tue, 27 Mar 2018 13:30:39 +0100 Subject: Comment --- synapse/storage/events.py | 9 +++++---- 1 file changed, 5 insertions(+), 4 deletions(-) (limited to 'synapse/storage/events.py') diff --git a/synapse/storage/events.py b/synapse/storage/events.py index 95b36d669b..f3d65f4338 100644 --- a/synapse/storage/events.py +++ b/synapse/storage/events.py @@ -415,10 +415,11 @@ class EventsStore(EventsWorkerStore): # processing one of these events. # What we're interested in is if the latest extremities # were the same when we created the event as they are - # now. We guess this by looking at the prev events and - # checking if they match up, as when this server creates - # a new event it will use the extremities as the prev - # events. + # now. When this server creates a new event (as opposed + # to receiving it over federation) it will use the + # forward extremities as the prev_events, so we can + # guess this by looking at the prev_events and checking + # if they match the current forward extremities. for ev, _ in ev_ctx_rm: prev_event_ids = set(e for e, _ in ev.prev_events) if latest_event_ids == prev_event_ids: -- cgit 1.5.1 From 05630758f25d958bf60fde4df5f80a89e4a9a0ac Mon Sep 17 00:00:00 2001 From: Richard van der Hoff Date: Thu, 29 Mar 2018 22:57:28 +0100 Subject: Use static JSONEncoders using json.dumps with custom options requires us to create a new JSONEncoder on each call. It's more efficient to create one upfront and reuse it. --- synapse/handlers/message.py | 4 ++-- synapse/replication/tcp/commands.py | 8 +++++--- synapse/storage/events.py | 23 ++++++++--------------- synapse/util/frozenutils.py | 19 +++++++++++++++++++ 4 files changed, 34 insertions(+), 20 deletions(-) (limited to 'synapse/storage/events.py') diff --git a/synapse/handlers/message.py b/synapse/handlers/message.py index 5a8ddc253e..6de6e13b7b 100644 --- a/synapse/handlers/message.py +++ b/synapse/handlers/message.py @@ -27,7 +27,7 @@ from synapse.types import ( from synapse.util.async import run_on_reactor, ReadWriteLock, Limiter from synapse.util.logcontext import preserve_fn, run_in_background from synapse.util.metrics import measure_func -from synapse.util.frozenutils import unfreeze +from synapse.util.frozenutils import frozendict_json_encoder from synapse.util.stringutils import random_string from synapse.visibility import filter_events_for_client from synapse.replication.http.send_event import send_event_to_master @@ -678,7 +678,7 @@ class EventCreationHandler(object): # Ensure that we can round trip before trying to persist in db try: - dump = simplejson.dumps(unfreeze(event.content)) + dump = frozendict_json_encoder.encode(event.content) simplejson.loads(dump) except Exception: logger.exception("Failed to encode content: %r", event.content) diff --git a/synapse/replication/tcp/commands.py b/synapse/replication/tcp/commands.py index 0005ad5879..34bcf903a3 100644 --- a/synapse/replication/tcp/commands.py +++ b/synapse/replication/tcp/commands.py @@ -24,6 +24,8 @@ import simplejson logger = logging.getLogger(__name__) +_json_encoder = simplejson.JSONEncoder(namedtuple_as_object=False) + class Command(object): """The base command class. @@ -107,7 +109,7 @@ class RdataCommand(Command): return " ".join(( self.stream_name, str(self.token) if self.token is not None else "batch", - simplejson.dumps(self.row, namedtuple_as_object=False), + _json_encoder.dumps(self.row), )) @@ -302,7 +304,7 @@ class InvalidateCacheCommand(Command): def to_line(self): return " ".join(( - self.cache_func, simplejson.dumps(self.keys, namedtuple_as_object=False) + self.cache_func, _json_encoder.encode(self.keys), )) @@ -334,7 +336,7 @@ class UserIpCommand(Command): ) def to_line(self): - return self.user_id + " " + simplejson.dumps(( + return self.user_id + " " + _json_encoder.encode(( self.access_token, self.ip, self.user_agent, self.device_id, self.last_seen, )) diff --git a/synapse/storage/events.py b/synapse/storage/events.py index f3d65f4338..ece5e6c41f 100644 --- a/synapse/storage/events.py +++ b/synapse/storage/events.py @@ -14,15 +14,19 @@ # See the License for the specific language governing permissions and # limitations under the License. -from synapse.storage.events_worker import EventsWorkerStore +from collections import OrderedDict, deque, namedtuple +from functools import wraps +import logging +import simplejson as json from twisted.internet import defer -from synapse.events import USE_FROZEN_DICTS +from synapse.storage.events_worker import EventsWorkerStore from synapse.util.async import ObservableDeferred +from synapse.util.frozenutils import frozendict_json_encoder from synapse.util.logcontext import ( - PreserveLoggingContext, make_deferred_yieldable + PreserveLoggingContext, make_deferred_yieldable, ) from synapse.util.logutils import log_function from synapse.util.metrics import Measure @@ -30,16 +34,8 @@ from synapse.api.constants import EventTypes from synapse.api.errors import SynapseError from synapse.util.caches.descriptors import cached, cachedInlineCallbacks from synapse.types import get_domain_from_id - -from canonicaljson import encode_canonical_json -from collections import deque, namedtuple, OrderedDict -from functools import wraps - import synapse.metrics -import logging -import simplejson as json - # these are only included to make the type annotations work from synapse.events import EventBase # noqa: F401 from synapse.events.snapshot import EventContext # noqa: F401 @@ -71,10 +67,7 @@ state_delta_reuse_delta_counter = metrics.register_counter( def encode_json(json_object): - if USE_FROZEN_DICTS: - return encode_canonical_json(json_object) - else: - return json.dumps(json_object, ensure_ascii=False) + return frozendict_json_encoder.encode(json_object) class _EventPeristenceQueue(object): diff --git a/synapse/util/frozenutils.py b/synapse/util/frozenutils.py index 6322f0f55c..f497b51f4a 100644 --- a/synapse/util/frozenutils.py +++ b/synapse/util/frozenutils.py @@ -14,6 +14,7 @@ # limitations under the License. from frozendict import frozendict +import simplejson as json def freeze(o): @@ -49,3 +50,21 @@ def unfreeze(o): pass return o + + +def _handle_frozendict(obj): + """Helper for EventEncoder. Makes frozendicts serializable by returning + the underlying dict + """ + if type(obj) is frozendict: + # fishing the protected dict out of the object is a bit nasty, + # but we don't really want the overhead of copying the dict. + return obj._dict + raise TypeError('Object of type %s is not JSON serializable' % + obj.__class__.__name__) + + +# A JSONEncoder which is capable of encoding frozendics without barfing +frozendict_json_encoder = json.JSONEncoder( + default=_handle_frozendict, +) -- cgit 1.5.1