From d90e8ea444092f500c721567fe055d48eb110528 Mon Sep 17 00:00:00 2001 From: Richard van der Hoff Date: Tue, 9 Jan 2018 18:27:35 +0000 Subject: Update http request metrics before calling servlet Make sure that we set the servlet name in the metrics object *before* calling the servlet, in case the servlet throws an exception. --- synapse/http/server.py | 30 +++++++++++++++++++----------- 1 file changed, 19 insertions(+), 11 deletions(-) (limited to 'synapse/http') diff --git a/synapse/http/server.py b/synapse/http/server.py index 25466cd292..6e8f4c9c5f 100644 --- a/synapse/http/server.py +++ b/synapse/http/server.py @@ -107,6 +107,10 @@ def wrap_request_handler(request_handler, include_metrics=False): with LoggingContext(request_id) as request_context: with Measure(self.clock, "wrapped_request_handler"): request_metrics = RequestMetrics() + # we start the request metrics timer here with an initial stab + # at the servlet name. For most requests that name will be + # JsonResource (or a subclass), and JsonResource._async_render + # will update it once it picks a servlet. request_metrics.start(self.clock, name=self.__class__.__name__) request_context.request = request_id @@ -249,12 +253,23 @@ class JsonResource(HttpServer, resource.Resource): if not m: continue - # We found a match! Trigger callback and then return the - # returned response. We pass both the request and any - # matched groups from the regex to the callback. + # We found a match! First update the metrics object to indicate + # which servlet is handling the request. callback = path_entry.callback + servlet_instance = getattr(callback, "__self__", None) + if servlet_instance is not None: + servlet_classname = servlet_instance.__class__.__name__ + else: + servlet_classname = "%r" % callback + + request_metrics.name = servlet_classname + + # Now trigger the callback. If it returns a response, we send it + # here. If it throws an exception, that is handled by the wrapper + # installed by @request_handler. + kwargs = intern_dict({ name: urllib.unquote(value).decode("UTF-8") if value else value for name, value in m.groupdict().items() @@ -265,17 +280,10 @@ class JsonResource(HttpServer, resource.Resource): code, response = callback_return self._send_response(request, code, response) - servlet_instance = getattr(callback, "__self__", None) - if servlet_instance is not None: - servlet_classname = servlet_instance.__class__.__name__ - else: - servlet_classname = "%r" % callback - - request_metrics.name = servlet_classname - return # Huh. No one wanted to handle that? Fiiiiiine. Send 400. + request_metrics.name = self.__class__.__name__ + ".UnrecognizedRequest" raise UnrecognizedRequestError() def _send_response(self, request, code, response_json_object, -- cgit 1.5.1 From 39f4e29d0151b56a3c8528e3149cd5765b9f600d Mon Sep 17 00:00:00 2001 From: Richard van der Hoff Date: Mon, 15 Jan 2018 17:00:12 +0000 Subject: Reorganise request and block metrics In order to circumvent the number of duplicate foo:count metrics increasing without bounds, it's time for a rearrangement. The following are all deprecated, and replaced with synapse_util_metrics_block_count: synapse_util_metrics_block_timer:count synapse_util_metrics_block_ru_utime:count synapse_util_metrics_block_ru_stime:count synapse_util_metrics_block_db_txn_count:count synapse_util_metrics_block_db_txn_duration:count The following are all deprecated, and replaced with synapse_http_server_response_count: synapse_http_server_requests synapse_http_server_response_time:count synapse_http_server_response_ru_utime:count synapse_http_server_response_ru_stime:count synapse_http_server_response_db_txn_count:count synapse_http_server_response_db_txn_duration:count The following are renamed (the old metrics are kept for now, but deprecated): synapse_util_metrics_block_timer:total -> synapse_util_metrics_block_time_seconds synapse_util_metrics_block_ru_utime:total -> synapse_util_metrics_block_ru_utime_seconds synapse_util_metrics_block_ru_stime:total -> synapse_util_metrics_block_ru_stime_seconds synapse_util_metrics_block_db_txn_count:total -> synapse_util_metrics_block_db_txn_count synapse_util_metrics_block_db_txn_duration:total -> synapse_util_metrics_block_db_txn_duration_seconds synapse_http_server_response_time:total -> synapse_http_server_response_time_seconds synapse_http_server_response_ru_utime:total -> synapse_http_server_response_ru_utime_seconds synapse_http_server_response_ru_stime:total -> synapse_http_server_response_ru_stime_seconds synapse_http_server_response_db_txn_count:total -> synapse_http_server_response_db_txn_count synapse_http_server_response_db_txn_duration:total synapse_http_server_response_db_txn_duration_seconds --- synapse/http/server.py | 56 ++++++++++++++++++++++++++++++++++++------------- synapse/util/metrics.py | 53 ++++++++++++++++++++++++++++++++++++---------- 2 files changed, 84 insertions(+), 25 deletions(-) (limited to 'synapse/http') diff --git a/synapse/http/server.py b/synapse/http/server.py index 6e8f4c9c5f..269b65ca41 100644 --- a/synapse/http/server.py +++ b/synapse/http/server.py @@ -42,34 +42,62 @@ logger = logging.getLogger(__name__) metrics = synapse.metrics.get_metrics_for(__name__) -incoming_requests_counter = metrics.register_counter( - "requests", +# total number of responses served, split by method/servlet/tag +response_count = metrics.register_counter( + "response_count", labels=["method", "servlet", "tag"], + alternative_names=( + # the following are all deprecated aliases for the same metric + metrics.name_prefix + x for x in ( + "_requests", + "_response_time:count", + "_response_ru_utime:count", + "_response_ru_stime:count", + "_response_db_txn_count:count", + "_response_db_txn_duration:count", + ) + ) ) + outgoing_responses_counter = metrics.register_counter( "responses", labels=["method", "code"], ) -response_timer = metrics.register_distribution( - "response_time", - labels=["method", "servlet", "tag"] +response_timer = metrics.register_counter( + "response_time_seconds", + labels=["method", "servlet", "tag"], + alternative_names=( + metrics.name_prefix + "_response_time:total", + ), ) -response_ru_utime = metrics.register_distribution( - "response_ru_utime", labels=["method", "servlet", "tag"] +response_ru_utime = metrics.register_counter( + "response_ru_utime_seconds", labels=["method", "servlet", "tag"], + alternative_names=( + metrics.name_prefix + "_response_ru_utime:total", + ), ) -response_ru_stime = metrics.register_distribution( - "response_ru_stime", labels=["method", "servlet", "tag"] +response_ru_stime = metrics.register_counter( + "response_ru_stime_seconds", labels=["method", "servlet", "tag"], + alternative_names=( + metrics.name_prefix + "_response_ru_stime:total", + ), ) -response_db_txn_count = metrics.register_distribution( - "response_db_txn_count", labels=["method", "servlet", "tag"] +response_db_txn_count = metrics.register_counter( + "response_db_txn_count", labels=["method", "servlet", "tag"], + alternative_names=( + metrics.name_prefix + "_response_db_txn_count:total", + ), ) -response_db_txn_duration = metrics.register_distribution( - "response_db_txn_duration", labels=["method", "servlet", "tag"] +response_db_txn_duration = metrics.register_counter( + "response_db_txn_duration_seconds", labels=["method", "servlet", "tag"], + alternative_names=( + metrics.name_prefix + "_response_db_txn_duration:total", + ), ) @@ -330,7 +358,7 @@ class RequestMetrics(object): ) return - incoming_requests_counter.inc(request.method, self.name, tag) + response_count.inc(request.method, self.name, tag) response_timer.inc_by( clock.time_msec() - self.start, request.method, diff --git a/synapse/util/metrics.py b/synapse/util/metrics.py index 4ea930d3e8..8d22ff3068 100644 --- a/synapse/util/metrics.py +++ b/synapse/util/metrics.py @@ -27,25 +27,56 @@ logger = logging.getLogger(__name__) metrics = synapse.metrics.get_metrics_for(__name__) -block_timer = metrics.register_distribution( - "block_timer", - labels=["block_name"] +# total number of times we have hit this block +response_count = metrics.register_counter( + "block_count", + labels=["block_name"], + alternative_names=( + # the following are all deprecated aliases for the same metric + metrics.name_prefix + x for x in ( + "_block_timer:count", + "_block_ru_utime:count", + "_block_ru_stime:count", + "_block_db_txn_count:count", + "_block_db_txn_duration:count", + ) + ) +) + +block_timer = metrics.register_counter( + "block_time_seconds", + labels=["block_name"], + alternative_names=( + metrics.name_prefix + "_block_timer:total", + ), ) -block_ru_utime = metrics.register_distribution( - "block_ru_utime", labels=["block_name"] +block_ru_utime = metrics.register_counter( + "block_ru_utime_seconds", labels=["block_name"], + alternative_names=( + metrics.name_prefix + "_block_ru_utime:total", + ), ) -block_ru_stime = metrics.register_distribution( - "block_ru_stime", labels=["block_name"] +block_ru_stime = metrics.register_counter( + "block_ru_stime_seconds", labels=["block_name"], + alternative_names=( + metrics.name_prefix + "_block_ru_stime:total", + ), ) -block_db_txn_count = metrics.register_distribution( - "block_db_txn_count", labels=["block_name"] +block_db_txn_count = metrics.register_counter( + "block_db_txn_count", labels=["block_name"], + alternative_names=( + metrics.name_prefix + "_block_db_txn_count:total", + ), ) -block_db_txn_duration = metrics.register_distribution( - "block_db_txn_duration", labels=["block_name"] +block_db_txn_duration = metrics.register_counter( + "block_db_txn_duration_seconds", labels=["block_name"], + alternative_names=( + metrics.name_prefix + "_block_db_txn_count:total", + ), ) -- cgit 1.5.1 From 6324b65f08b3f8dbfee6fef0079e2a87cb1c2c85 Mon Sep 17 00:00:00 2001 From: Richard van der Hoff Date: Thu, 11 Jan 2018 18:17:54 +0000 Subject: Track db txn time in millisecs ... to reduce the amount of floating-point foo we do. --- synapse/http/server.py | 4 +++- synapse/http/site.py | 6 +++--- synapse/util/logcontext.py | 9 ++++++--- synapse/util/metrics.py | 8 +++++--- 4 files changed, 17 insertions(+), 10 deletions(-) (limited to 'synapse/http') diff --git a/synapse/http/server.py b/synapse/http/server.py index 269b65ca41..0f30e6fd56 100644 --- a/synapse/http/server.py +++ b/synapse/http/server.py @@ -93,6 +93,8 @@ response_db_txn_count = metrics.register_counter( ), ) +# seconds spent waiting for db txns, excluding scheduling time, when processing +# this request response_db_txn_duration = metrics.register_counter( "response_db_txn_duration_seconds", labels=["method", "servlet", "tag"], alternative_names=( @@ -377,7 +379,7 @@ class RequestMetrics(object): context.db_txn_count, request.method, self.name, tag ) response_db_txn_duration.inc_by( - context.db_txn_duration, request.method, self.name, tag + context.db_txn_duration_ms / 1000., request.method, self.name, tag ) diff --git a/synapse/http/site.py b/synapse/http/site.py index cd1492b1c3..dc64f0f6f5 100644 --- a/synapse/http/site.py +++ b/synapse/http/site.py @@ -66,10 +66,10 @@ class SynapseRequest(Request): context = LoggingContext.current_context() ru_utime, ru_stime = context.get_resource_usage() db_txn_count = context.db_txn_count - db_txn_duration = context.db_txn_duration + db_txn_duration_ms = context.db_txn_duration_ms except Exception: ru_utime, ru_stime = (0, 0) - db_txn_count, db_txn_duration = (0, 0) + db_txn_count, db_txn_duration_ms = (0, 0) self.site.access_logger.info( "%s - %s - {%s}" @@ -81,7 +81,7 @@ class SynapseRequest(Request): int(time.time() * 1000) - self.start_time, int(ru_utime * 1000), int(ru_stime * 1000), - int(db_txn_duration * 1000), + db_txn_duration_ms, int(db_txn_count), self.sentLength, self.code, diff --git a/synapse/util/logcontext.py b/synapse/util/logcontext.py index ca71a1fc27..a78e53812f 100644 --- a/synapse/util/logcontext.py +++ b/synapse/util/logcontext.py @@ -59,7 +59,7 @@ class LoggingContext(object): __slots__ = [ "previous_context", "name", "ru_stime", "ru_utime", - "db_txn_count", "db_txn_duration", "usage_start", "usage_end", + "db_txn_count", "db_txn_duration_ms", "usage_start", "usage_end", "main_thread", "alive", "request", "tag", ] @@ -97,7 +97,10 @@ class LoggingContext(object): self.ru_stime = 0. self.ru_utime = 0. self.db_txn_count = 0 - self.db_txn_duration = 0. + + # ms spent waiting for db txns, excluding scheduling time + self.db_txn_duration_ms = 0 + self.usage_start = None self.usage_end = None self.main_thread = threading.current_thread() @@ -205,7 +208,7 @@ class LoggingContext(object): def add_database_transaction(self, duration_ms): self.db_txn_count += 1 - self.db_txn_duration += duration_ms / 1000. + self.db_txn_duration_ms += duration_ms class LoggingContextFilter(logging.Filter): diff --git a/synapse/util/metrics.py b/synapse/util/metrics.py index 8d22ff3068..d25629cc50 100644 --- a/synapse/util/metrics.py +++ b/synapse/util/metrics.py @@ -72,6 +72,7 @@ block_db_txn_count = metrics.register_counter( ), ) +# seconds spent waiting for db txns, excluding scheduling time, in this block block_db_txn_duration = metrics.register_counter( "block_db_txn_duration_seconds", labels=["block_name"], alternative_names=( @@ -95,7 +96,7 @@ def measure_func(name): class Measure(object): __slots__ = [ "clock", "name", "start_context", "start", "new_context", "ru_utime", - "ru_stime", "db_txn_count", "db_txn_duration", "created_context" + "ru_stime", "db_txn_count", "db_txn_duration_ms", "created_context" ] def __init__(self, clock, name): @@ -115,7 +116,7 @@ class Measure(object): self.ru_utime, self.ru_stime = self.start_context.get_resource_usage() self.db_txn_count = self.start_context.db_txn_count - self.db_txn_duration = self.start_context.db_txn_duration + self.db_txn_duration_ms = self.start_context.db_txn_duration_ms def __exit__(self, exc_type, exc_val, exc_tb): if isinstance(exc_type, Exception) or not self.start_context: @@ -145,7 +146,8 @@ class Measure(object): context.db_txn_count - self.db_txn_count, self.name ) block_db_txn_duration.inc_by( - context.db_txn_duration - self.db_txn_duration, self.name + (context.db_txn_duration_ms - self.db_txn_duration_ms) / 1000., + self.name ) if self.created_context: -- cgit 1.5.1 From 3d12d97415ac6d6a4ab8188af31c7df12c5d19f8 Mon Sep 17 00:00:00 2001 From: Richard van der Hoff Date: Fri, 12 Jan 2018 00:27:14 +0000 Subject: Track DB scheduling delay per-request For each request, track the amount of time spent waiting for a db connection. This entails adding it to the LoggingContext and we may as well add metrics for it while we are passing. --- synapse/http/server.py | 7 +++++++ synapse/http/site.py | 4 +++- synapse/storage/_base.py | 4 +++- synapse/util/logcontext.py | 18 +++++++++++++++++- synapse/util/metrics.py | 14 +++++++++++++- 5 files changed, 43 insertions(+), 4 deletions(-) (limited to 'synapse/http') diff --git a/synapse/http/server.py b/synapse/http/server.py index 0f30e6fd56..7b6418bc2c 100644 --- a/synapse/http/server.py +++ b/synapse/http/server.py @@ -102,6 +102,10 @@ response_db_txn_duration = metrics.register_counter( ), ) +# seconds spent waiting for a db connection, when processing this request +response_db_sched_duration = metrics.register_counter( + "response_db_sched_duration_seconds", labels=["method", "servlet", "tag"] +) _next_request_id = 0 @@ -381,6 +385,9 @@ class RequestMetrics(object): response_db_txn_duration.inc_by( context.db_txn_duration_ms / 1000., request.method, self.name, tag ) + response_db_sched_duration.inc_by( + context.db_sched_duration_ms / 1000., request.method, self.name, tag + ) class RootRedirect(resource.Resource): diff --git a/synapse/http/site.py b/synapse/http/site.py index dc64f0f6f5..e422c8dfae 100644 --- a/synapse/http/site.py +++ b/synapse/http/site.py @@ -67,13 +67,14 @@ class SynapseRequest(Request): ru_utime, ru_stime = context.get_resource_usage() db_txn_count = context.db_txn_count db_txn_duration_ms = context.db_txn_duration_ms + db_sched_duration_ms = context.db_sched_duration_ms except Exception: ru_utime, ru_stime = (0, 0) db_txn_count, db_txn_duration_ms = (0, 0) self.site.access_logger.info( "%s - %s - {%s}" - " Processed request: %dms (%dms, %dms) (%dms/%d)" + " Processed request: %dms (%dms, %dms) (%dms/%dms/%d)" " %sB %s \"%s %s %s\" \"%s\"", self.getClientIP(), self.site.site_tag, @@ -81,6 +82,7 @@ class SynapseRequest(Request): int(time.time() * 1000) - self.start_time, int(ru_utime * 1000), int(ru_stime * 1000), + db_sched_duration_ms, db_txn_duration_ms, int(db_txn_count), self.sentLength, diff --git a/synapse/storage/_base.py b/synapse/storage/_base.py index 986617674c..68125006eb 100644 --- a/synapse/storage/_base.py +++ b/synapse/storage/_base.py @@ -347,7 +347,9 @@ class SQLBaseStore(object): def inner_func(conn, *args, **kwargs): with LoggingContext("runWithConnection") as context: - sql_scheduling_timer.inc_by(time.time() * 1000 - start_time) + sched_duration_ms = time.time() * 1000 - start_time + sql_scheduling_timer.inc_by(sched_duration_ms) + current_context.add_database_scheduled(sched_duration_ms) if self.database_engine.is_connection_closed(conn): logger.debug("Reconnecting closed database connection") diff --git a/synapse/util/logcontext.py b/synapse/util/logcontext.py index a78e53812f..94fa7cac98 100644 --- a/synapse/util/logcontext.py +++ b/synapse/util/logcontext.py @@ -59,7 +59,8 @@ class LoggingContext(object): __slots__ = [ "previous_context", "name", "ru_stime", "ru_utime", - "db_txn_count", "db_txn_duration_ms", "usage_start", "usage_end", + "db_txn_count", "db_txn_duration_ms", "db_sched_duration_ms", + "usage_start", "usage_end", "main_thread", "alive", "request", "tag", ] @@ -86,6 +87,9 @@ class LoggingContext(object): def add_database_transaction(self, duration_ms): pass + def add_database_scheduled(self, sched_ms): + pass + def __nonzero__(self): return False @@ -101,6 +105,9 @@ class LoggingContext(object): # ms spent waiting for db txns, excluding scheduling time self.db_txn_duration_ms = 0 + # ms spent waiting for db txns to be scheduled + self.db_sched_duration_ms = 0 + self.usage_start = None self.usage_end = None self.main_thread = threading.current_thread() @@ -210,6 +217,15 @@ class LoggingContext(object): self.db_txn_count += 1 self.db_txn_duration_ms += duration_ms + def add_database_scheduled(self, sched_ms): + """Record a use of the database pool + + Args: + sched_ms (int): number of milliseconds it took us to get a + connection + """ + self.db_sched_duration_ms += sched_ms + class LoggingContextFilter(logging.Filter): """Logging filter that adds values from the current logging context to each diff --git a/synapse/util/metrics.py b/synapse/util/metrics.py index d25629cc50..059bb7fedf 100644 --- a/synapse/util/metrics.py +++ b/synapse/util/metrics.py @@ -80,6 +80,11 @@ block_db_txn_duration = metrics.register_counter( ), ) +# seconds spent waiting for a db connection, in this block +block_db_sched_duration = metrics.register_counter( + "block_db_sched_duration_seconds", labels=["block_name"], +) + def measure_func(name): def wrapper(func): @@ -96,7 +101,9 @@ def measure_func(name): class Measure(object): __slots__ = [ "clock", "name", "start_context", "start", "new_context", "ru_utime", - "ru_stime", "db_txn_count", "db_txn_duration_ms", "created_context" + "ru_stime", + "db_txn_count", "db_txn_duration_ms", "db_sched_duration_ms", + "created_context", ] def __init__(self, clock, name): @@ -117,6 +124,7 @@ class Measure(object): self.ru_utime, self.ru_stime = self.start_context.get_resource_usage() self.db_txn_count = self.start_context.db_txn_count self.db_txn_duration_ms = self.start_context.db_txn_duration_ms + self.db_sched_duration_ms = self.start_context.db_sched_duration_ms def __exit__(self, exc_type, exc_val, exc_tb): if isinstance(exc_type, Exception) or not self.start_context: @@ -149,6 +157,10 @@ class Measure(object): (context.db_txn_duration_ms - self.db_txn_duration_ms) / 1000., self.name ) + block_db_sched_duration.inc_by( + (context.db_sched_duration_ms - self.db_sched_duration_ms) / 1000., + self.name + ) if self.created_context: self.start_context.__exit__(exc_type, exc_val, exc_tb) -- cgit 1.5.1 From 936482d507ffbf59d9aee58f851ec6ff2a120424 Mon Sep 17 00:00:00 2001 From: Richard van der Hoff Date: Tue, 16 Jan 2018 17:58:16 +0000 Subject: Fix 'NoneType' object has no attribute 'writeHeaders' Avoid throwing a (harmless) exception when we try to write an error response to an http request where the client has disconnected. This comes up as a CRITICAL error in the logs which tends to mislead people into thinking there's an actual problem --- synapse/http/server.py | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-) (limited to 'synapse/http') diff --git a/synapse/http/server.py b/synapse/http/server.py index 269b65ca41..8077e22794 100644 --- a/synapse/http/server.py +++ b/synapse/http/server.py @@ -316,15 +316,6 @@ class JsonResource(HttpServer, resource.Resource): def _send_response(self, request, code, response_json_object, response_code_message=None): - # could alternatively use request.notifyFinish() and flip a flag when - # the Deferred fires, but since the flag is RIGHT THERE it seems like - # a waste. - if request._disconnected: - logger.warn( - "Not sending response to request %s, already disconnected.", - request) - return - outgoing_responses_counter.inc(request.method, str(code)) # TODO: Only enable CORS for the requests that need it. @@ -400,6 +391,15 @@ class RootRedirect(resource.Resource): def respond_with_json(request, code, json_object, send_cors=False, response_code_message=None, pretty_print=False, version_string="", canonical_json=True): + # could alternatively use request.notifyFinish() and flip a flag when + # the Deferred fires, but since the flag is RIGHT THERE it seems like + # a waste. + if request._disconnected: + logger.warn( + "Not sending response to request %s, already disconnected.", + request) + return + if pretty_print: json_bytes = encode_pretty_printed_json(json_object) + "\n" else: -- cgit 1.5.1 From 2c8526cac795fa2aa795e1a1aaae2ffb2558824d Mon Sep 17 00:00:00 2001 From: Richard van der Hoff Date: Sat, 20 Jan 2018 00:55:44 +0000 Subject: Use a connection pool for the SimpleHttpClient In particular I hope this will help the pusher, which makes many requests to sygnal, and is currently negotiating SSL for each one. --- synapse/http/client.py | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) (limited to 'synapse/http') diff --git a/synapse/http/client.py b/synapse/http/client.py index 4abb479ae3..930d713013 100644 --- a/synapse/http/client.py +++ b/synapse/http/client.py @@ -30,6 +30,7 @@ from twisted.internet.endpoints import HostnameEndpoint, wrapClientTLS from twisted.web.client import ( BrowserLikeRedirectAgent, ContentDecoderAgent, GzipDecoder, Agent, readBody, PartialDownloadError, + HTTPConnectionPool, ) from twisted.web.client import FileBodyProducer as TwistedFileBodyProducer from twisted.web.http import PotentialDataLoss @@ -64,13 +65,19 @@ class SimpleHttpClient(object): """ def __init__(self, hs): self.hs = hs + + pool = HTTPConnectionPool(reactor) + pool.maxPersistentPerHost = 5 + pool.cachedConnectionTimeout = 2 * 60 + # The default context factory in Twisted 14.0.0 (which we require) is # BrowserLikePolicyForHTTPS which will do regular cert validation # 'like a browser' self.agent = Agent( reactor, connectTimeout=15, - contextFactory=hs.get_http_client_context_factory() + contextFactory=hs.get_http_client_context_factory(), + pool=pool, ) self.user_agent = hs.version_string self.clock = hs.get_clock() -- cgit 1.5.1 From ab9f844aaf3662a64dbc4c56077e9fa37bc7d5d0 Mon Sep 17 00:00:00 2001 From: Matthew Hodgson Date: Mon, 22 Jan 2018 19:11:18 +0100 Subject: Add federation_domain_whitelist option (#2820) Add federation_domain_whitelist gives a way to restrict which domains your HS is allowed to federate with. useful mainly for gracefully preventing a private but internet-connected HS from trying to federate to the wider public Matrix network --- synapse/api/errors.py | 26 ++++++++++++++++++++++++++ synapse/config/server.py | 22 ++++++++++++++++++++++ synapse/federation/federation_client.py | 5 ++++- synapse/federation/transaction_queue.py | 4 +++- synapse/federation/transport/client.py | 3 +++ synapse/federation/transport/server.py | 9 ++++++++- synapse/handlers/device.py | 4 ++++ synapse/handlers/e2e_keys.py | 8 +++++++- synapse/handlers/federation.py | 4 ++++ synapse/http/matrixfederationclient.py | 28 +++++++++++++++++++++++++++- synapse/rest/key/v2/remote_key_resource.py | 8 ++++++++ synapse/rest/media/v1/media_repository.py | 19 +++++++++++++++++-- synapse/util/retryutils.py | 12 ++++++++++++ tests/utils.py | 1 + 14 files changed, 146 insertions(+), 7 deletions(-) (limited to 'synapse/http') diff --git a/synapse/api/errors.py b/synapse/api/errors.py index 46b0d7b34c..aa15f73f36 100644 --- a/synapse/api/errors.py +++ b/synapse/api/errors.py @@ -141,6 +141,32 @@ class RegistrationError(SynapseError): pass +class FederationDeniedError(SynapseError): + """An error raised when the server tries to federate with a server which + is not on its federation whitelist. + + Attributes: + destination (str): The destination which has been denied + """ + + def __init__(self, destination): + """Raised by federation client or server to indicate that we are + are deliberately not attempting to contact a given server because it is + not on our federation whitelist. + + Args: + destination (str): the domain in question + """ + + self.destination = destination + + super(FederationDeniedError, self).__init__( + code=403, + msg="Federation denied with %s." % (self.destination,), + errcode=Codes.FORBIDDEN, + ) + + class InteractiveAuthIncompleteError(Exception): """An error raised when UI auth is not yet complete diff --git a/synapse/config/server.py b/synapse/config/server.py index 436dd8a6fe..8f0b6d1f28 100644 --- a/synapse/config/server.py +++ b/synapse/config/server.py @@ -55,6 +55,17 @@ class ServerConfig(Config): "block_non_admin_invites", False, ) + # FIXME: federation_domain_whitelist needs sytests + self.federation_domain_whitelist = None + federation_domain_whitelist = config.get( + "federation_domain_whitelist", None + ) + # turn the whitelist into a hash for speed of lookup + if federation_domain_whitelist is not None: + self.federation_domain_whitelist = {} + for domain in federation_domain_whitelist: + self.federation_domain_whitelist[domain] = True + if self.public_baseurl is not None: if self.public_baseurl[-1] != '/': self.public_baseurl += '/' @@ -210,6 +221,17 @@ class ServerConfig(Config): # (except those sent by local server admins). The default is False. # block_non_admin_invites: True + # Restrict federation to the following whitelist of domains. + # N.B. we recommend also firewalling your federation listener to limit + # inbound federation traffic as early as possible, rather than relying + # purely on this application-layer restriction. If not specified, the + # default is to whitelist everything. + # + # federation_domain_whitelist: + # - lon.example.com + # - nyc.example.com + # - syd.example.com + # List of ports that Synapse should listen on, their purpose and their # configuration. listeners: diff --git a/synapse/federation/federation_client.py b/synapse/federation/federation_client.py index b1fe03f702..813907f7f2 100644 --- a/synapse/federation/federation_client.py +++ b/synapse/federation/federation_client.py @@ -23,7 +23,7 @@ from twisted.internet import defer from synapse.api.constants import Membership from synapse.api.errors import ( - CodeMessageException, HttpResponseException, SynapseError, + CodeMessageException, HttpResponseException, SynapseError, FederationDeniedError ) from synapse.events import builder from synapse.federation.federation_base import ( @@ -266,6 +266,9 @@ class FederationClient(FederationBase): except NotRetryingDestination as e: logger.info(e.message) continue + except FederationDeniedError as e: + logger.info(e.message) + continue except Exception as e: pdu_attempts[destination] = now diff --git a/synapse/federation/transaction_queue.py b/synapse/federation/transaction_queue.py index 9d39f46583..a141ec9953 100644 --- a/synapse/federation/transaction_queue.py +++ b/synapse/federation/transaction_queue.py @@ -19,7 +19,7 @@ from twisted.internet import defer from .persistence import TransactionActions from .units import Transaction, Edu -from synapse.api.errors import HttpResponseException +from synapse.api.errors import HttpResponseException, FederationDeniedError from synapse.util import logcontext, PreserveLoggingContext from synapse.util.async import run_on_reactor from synapse.util.retryutils import NotRetryingDestination, get_retry_limiter @@ -490,6 +490,8 @@ class TransactionQueue(object): (e.retry_last_ts + e.retry_interval) / 1000.0 ), ) + except FederationDeniedError as e: + logger.info(e) except Exception as e: logger.warn( "TX [%s] Failed to send transaction: %s", diff --git a/synapse/federation/transport/client.py b/synapse/federation/transport/client.py index 1f3ce238f6..5488e82985 100644 --- a/synapse/federation/transport/client.py +++ b/synapse/federation/transport/client.py @@ -212,6 +212,9 @@ class TransportLayerClient(object): Fails with ``NotRetryingDestination`` if we are not yet ready to retry this server. + + Fails with ``FederationDeniedError`` if the remote destination + is not in our federation whitelist """ valid_memberships = {Membership.JOIN, Membership.LEAVE} if membership not in valid_memberships: diff --git a/synapse/federation/transport/server.py b/synapse/federation/transport/server.py index 2b02b021ec..06c16ba4fa 100644 --- a/synapse/federation/transport/server.py +++ b/synapse/federation/transport/server.py @@ -16,7 +16,7 @@ from twisted.internet import defer from synapse.api.urls import FEDERATION_PREFIX as PREFIX -from synapse.api.errors import Codes, SynapseError +from synapse.api.errors import Codes, SynapseError, FederationDeniedError from synapse.http.server import JsonResource from synapse.http.servlet import ( parse_json_object_from_request, parse_integer_from_args, parse_string_from_args, @@ -81,6 +81,7 @@ class Authenticator(object): self.keyring = hs.get_keyring() self.server_name = hs.hostname self.store = hs.get_datastore() + self.federation_domain_whitelist = hs.config.federation_domain_whitelist # A method just so we can pass 'self' as the authenticator to the Servlets @defer.inlineCallbacks @@ -92,6 +93,12 @@ class Authenticator(object): "signatures": {}, } + if ( + self.federation_domain_whitelist is not None and + self.server_name not in self.federation_domain_whitelist + ): + raise FederationDeniedError(self.server_name) + if content is not None: json_request["content"] = content diff --git a/synapse/handlers/device.py b/synapse/handlers/device.py index 2152efc692..0e83453851 100644 --- a/synapse/handlers/device.py +++ b/synapse/handlers/device.py @@ -14,6 +14,7 @@ # limitations under the License. from synapse.api import errors from synapse.api.constants import EventTypes +from synapse.api.errors import FederationDeniedError from synapse.util import stringutils from synapse.util.async import Linearizer from synapse.util.caches.expiringcache import ExpiringCache @@ -513,6 +514,9 @@ class DeviceListEduUpdater(object): # This makes it more likely that the device lists will # eventually become consistent. return + except FederationDeniedError as e: + logger.info(e) + return except Exception: # TODO: Remember that we are now out of sync and try again # later diff --git a/synapse/handlers/e2e_keys.py b/synapse/handlers/e2e_keys.py index 5af8abf66b..9aa95f89e6 100644 --- a/synapse/handlers/e2e_keys.py +++ b/synapse/handlers/e2e_keys.py @@ -19,7 +19,9 @@ import logging from canonicaljson import encode_canonical_json from twisted.internet import defer -from synapse.api.errors import SynapseError, CodeMessageException +from synapse.api.errors import ( + SynapseError, CodeMessageException, FederationDeniedError, +) from synapse.types import get_domain_from_id, UserID from synapse.util.logcontext import preserve_fn, make_deferred_yieldable from synapse.util.retryutils import NotRetryingDestination @@ -140,6 +142,10 @@ class E2eKeysHandler(object): failures[destination] = { "status": 503, "message": "Not ready for retry", } + except FederationDeniedError as e: + failures[destination] = { + "status": 403, "message": "Federation Denied", + } except Exception as e: # include ConnectionRefused and other errors failures[destination] = { diff --git a/synapse/handlers/federation.py b/synapse/handlers/federation.py index ac70730885..677532c87b 100644 --- a/synapse/handlers/federation.py +++ b/synapse/handlers/federation.py @@ -22,6 +22,7 @@ from ._base import BaseHandler from synapse.api.errors import ( AuthError, FederationError, StoreError, CodeMessageException, SynapseError, + FederationDeniedError, ) from synapse.api.constants import EventTypes, Membership, RejectedReason from synapse.events.validator import EventValidator @@ -782,6 +783,9 @@ class FederationHandler(BaseHandler): except NotRetryingDestination as e: logger.info(e.message) continue + except FederationDeniedError as e: + logger.info(e) + continue except Exception as e: logger.exception( "Failed to backfill from %s because %s", diff --git a/synapse/http/matrixfederationclient.py b/synapse/http/matrixfederationclient.py index 833496b72d..9145405cb0 100644 --- a/synapse/http/matrixfederationclient.py +++ b/synapse/http/matrixfederationclient.py @@ -27,7 +27,7 @@ import synapse.metrics from canonicaljson import encode_canonical_json from synapse.api.errors import ( - SynapseError, Codes, HttpResponseException, + SynapseError, Codes, HttpResponseException, FederationDeniedError, ) from signedjson.sign import sign_json @@ -123,11 +123,22 @@ class MatrixFederationHttpClient(object): Fails with ``HTTPRequestException``: if we get an HTTP response code >= 300. + Fails with ``NotRetryingDestination`` if we are not yet ready to retry this server. + + Fails with ``FederationDeniedError`` if this destination + is not on our federation whitelist + (May also fail with plenty of other Exceptions for things like DNS failures, connection failures, SSL failures.) """ + if ( + self.hs.config.federation_domain_whitelist and + destination not in self.hs.config.federation_domain_whitelist + ): + raise FederationDeniedError(destination) + limiter = yield synapse.util.retryutils.get_retry_limiter( destination, self.clock, @@ -308,6 +319,9 @@ class MatrixFederationHttpClient(object): Fails with ``NotRetryingDestination`` if we are not yet ready to retry this server. + + Fails with ``FederationDeniedError`` if this destination + is not on our federation whitelist """ if not json_data_callback: @@ -368,6 +382,9 @@ class MatrixFederationHttpClient(object): Fails with ``NotRetryingDestination`` if we are not yet ready to retry this server. + + Fails with ``FederationDeniedError`` if this destination + is not on our federation whitelist """ def body_callback(method, url_bytes, headers_dict): @@ -422,6 +439,9 @@ class MatrixFederationHttpClient(object): Fails with ``NotRetryingDestination`` if we are not yet ready to retry this server. + + Fails with ``FederationDeniedError`` if this destination + is not on our federation whitelist """ logger.debug("get_json args: %s", args) @@ -475,6 +495,9 @@ class MatrixFederationHttpClient(object): Fails with ``NotRetryingDestination`` if we are not yet ready to retry this server. + + Fails with ``FederationDeniedError`` if this destination + is not on our federation whitelist """ response = yield self._request( @@ -518,6 +541,9 @@ class MatrixFederationHttpClient(object): Fails with ``NotRetryingDestination`` if we are not yet ready to retry this server. + + Fails with ``FederationDeniedError`` if this destination + is not on our federation whitelist """ encoded_args = {} diff --git a/synapse/rest/key/v2/remote_key_resource.py b/synapse/rest/key/v2/remote_key_resource.py index cc2842aa72..17e6079cba 100644 --- a/synapse/rest/key/v2/remote_key_resource.py +++ b/synapse/rest/key/v2/remote_key_resource.py @@ -93,6 +93,7 @@ class RemoteKey(Resource): self.store = hs.get_datastore() self.version_string = hs.version_string self.clock = hs.get_clock() + self.federation_domain_whitelist = hs.config.federation_domain_whitelist def render_GET(self, request): self.async_render_GET(request) @@ -137,6 +138,13 @@ class RemoteKey(Resource): logger.info("Handling query for keys %r", query) store_queries = [] for server_name, key_ids in query.items(): + if ( + self.federation_domain_whitelist is not None and + server_name not in self.federation_domain_whitelist + ): + logger.debug("Federation denied with %s", server_name) + continue + if not key_ids: key_ids = (None,) for key_id in key_ids: diff --git a/synapse/rest/media/v1/media_repository.py b/synapse/rest/media/v1/media_repository.py index 4f56bcf577..485db8577a 100644 --- a/synapse/rest/media/v1/media_repository.py +++ b/synapse/rest/media/v1/media_repository.py @@ -32,8 +32,9 @@ from .media_storage import MediaStorage from synapse.http.matrixfederationclient import MatrixFederationHttpClient from synapse.util.stringutils import random_string -from synapse.api.errors import SynapseError, HttpResponseException, \ - NotFoundError +from synapse.api.errors import ( + SynapseError, HttpResponseException, NotFoundError, FederationDeniedError, +) from synapse.util.async import Linearizer from synapse.util.stringutils import is_ascii @@ -75,6 +76,8 @@ class MediaRepository(object): self.recently_accessed_remotes = set() self.recently_accessed_locals = set() + self.federation_domain_whitelist = hs.config.federation_domain_whitelist + # List of StorageProviders where we should search for media and # potentially upload to. storage_providers = [] @@ -216,6 +219,12 @@ class MediaRepository(object): Deferred: Resolves once a response has successfully been written to request """ + if ( + self.federation_domain_whitelist is not None and + server_name not in self.federation_domain_whitelist + ): + raise FederationDeniedError(server_name) + self.mark_recently_accessed(server_name, media_id) # We linearize here to ensure that we don't try and download remote @@ -250,6 +259,12 @@ class MediaRepository(object): Returns: Deferred[dict]: The media_info of the file """ + if ( + self.federation_domain_whitelist is not None and + server_name not in self.federation_domain_whitelist + ): + raise FederationDeniedError(server_name) + # We linearize here to ensure that we don't try and download remote # media multiple times concurrently key = (server_name, media_id) diff --git a/synapse/util/retryutils.py b/synapse/util/retryutils.py index 1adedbb361..47b0bb5eb3 100644 --- a/synapse/util/retryutils.py +++ b/synapse/util/retryutils.py @@ -26,6 +26,18 @@ logger = logging.getLogger(__name__) class NotRetryingDestination(Exception): def __init__(self, retry_last_ts, retry_interval, destination): + """Raised by the limiter (and federation client) to indicate that we are + are deliberately not attempting to contact a given server. + + Args: + retry_last_ts (int): the unix ts in milliseconds of our last attempt + to contact the server. 0 indicates that the last attempt was + successful or that we've never actually attempted to connect. + retry_interval (int): the time in milliseconds to wait until the next + attempt. + destination (str): the domain in question + """ + msg = "Not retrying server %s." % (destination,) super(NotRetryingDestination, self).__init__(msg) diff --git a/tests/utils.py b/tests/utils.py index 44e5f75093..3116047892 100644 --- a/tests/utils.py +++ b/tests/utils.py @@ -57,6 +57,7 @@ def setup_test_homeserver(name="test", datastore=None, config=None, **kargs): config.worker_app = None config.email_enable_notifs = False config.block_non_admin_invites = False + config.federation_domain_whitelist = None # disable user directory updates, because they get done in the # background, which upsets the test runner. -- cgit 1.5.1 From 02ba118f81009b1c3ae290a17c35e1b9d75e802b Mon Sep 17 00:00:00 2001 From: Richard van der Hoff Date: Mon, 29 Jan 2018 14:30:15 +0000 Subject: Increase http conn pool size --- synapse/http/client.py | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) (limited to 'synapse/http') diff --git a/synapse/http/client.py b/synapse/http/client.py index 930d713013..f3e4973c2e 100644 --- a/synapse/http/client.py +++ b/synapse/http/client.py @@ -18,6 +18,7 @@ from OpenSSL.SSL import VERIFY_NONE from synapse.api.errors import ( CodeMessageException, MatrixCodeMessageException, SynapseError, Codes, ) +from synapse.util.caches import CACHE_SIZE_FACTOR from synapse.util.logcontext import make_deferred_yieldable from synapse.util import logcontext import synapse.metrics @@ -67,7 +68,11 @@ class SimpleHttpClient(object): self.hs = hs pool = HTTPConnectionPool(reactor) - pool.maxPersistentPerHost = 5 + + # the pusher makes lots of concurrent SSL connections to sygnal, and + # tends to do so in batches, so we need to allow the pool to keep lots + # of idle connections around. + pool.maxPersistentPerHost = max((100 * CACHE_SIZE_FACTOR, 5)) pool.cachedConnectionTimeout = 2 * 60 # The default context factory in Twisted 14.0.0 (which we require) is -- cgit 1.5.1 From af19f5e9aa47dd9875df3937d514ae7708eed539 Mon Sep 17 00:00:00 2001 From: Richard van der Hoff Date: Tue, 30 Jan 2018 17:52:03 +0000 Subject: Remove spurious log argument ... which would cause scary-looking and unhelpful errors in the log on dns fail --- synapse/http/endpoint.py | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) (limited to 'synapse/http') diff --git a/synapse/http/endpoint.py b/synapse/http/endpoint.py index e2b99ef3bd..87639b9151 100644 --- a/synapse/http/endpoint.py +++ b/synapse/http/endpoint.py @@ -357,8 +357,7 @@ def _get_hosts_for_srv_record(dns_client, host): def eb(res, record_type): if res.check(DNSNameError): return [] - logger.warn("Error looking up %s for %s: %s", - record_type, host, res, res.value) + logger.warn("Error looking up %s for %s: %s", record_type, host, res) return res # no logcontexts here, so we can safely fire these off and gatherResults -- cgit 1.5.1 From 74fcbf741b3a7b95b5cc44478050e8a40fb7dc46 Mon Sep 17 00:00:00 2001 From: Richard van der Hoff Date: Thu, 8 Feb 2018 18:44:52 +0000 Subject: delete_local_events for purge_history Add a flag which makes the purger delete local events --- docs/admin_api/purge_history_api.rst | 14 ++++++++++++-- synapse/handlers/message.py | 4 ++-- synapse/http/servlet.py | 18 +++++++++++++++--- synapse/rest/client/v1/admin.py | 11 ++++++++++- synapse/storage/events.py | 35 ++++++++++++++++++++++++++++------- 5 files changed, 67 insertions(+), 15 deletions(-) (limited to 'synapse/http') diff --git a/docs/admin_api/purge_history_api.rst b/docs/admin_api/purge_history_api.rst index 08b3306366..b4e5bd9d75 100644 --- a/docs/admin_api/purge_history_api.rst +++ b/docs/admin_api/purge_history_api.rst @@ -4,8 +4,6 @@ Purge History API The purge history API allows server admins to purge historic events from their database, reclaiming disk space. -**NB!** This will not delete local events (locally sent messages content etc) from the database, but will remove lots of the metadata about them and does dramatically reduce the on disk space usage - Depending on the amount of history being purged a call to the API may take several minutes or longer. During this period users will not be able to paginate further back in the room from the point being purged from. @@ -15,3 +13,15 @@ The API is simply: ``POST /_matrix/client/r0/admin/purge_history//`` including an ``access_token`` of a server admin. + +By default, events sent by local users are not deleted, as they may represent +the only copies of this content in existence. (Events sent by remote users are +deleted, and room state data before the cutoff is always removed). + +To delete local events as well, set ``delete_local_events`` in the body: + +.. code:: json + + { + "delete_local_events": True, + } diff --git a/synapse/handlers/message.py b/synapse/handlers/message.py index 1c7860bb05..276d1a7722 100644 --- a/synapse/handlers/message.py +++ b/synapse/handlers/message.py @@ -63,7 +63,7 @@ class MessageHandler(BaseHandler): self.spam_checker = hs.get_spam_checker() @defer.inlineCallbacks - def purge_history(self, room_id, event_id): + def purge_history(self, room_id, event_id, delete_local_events=False): event = yield self.store.get_event(event_id) if event.room_id != room_id: @@ -72,7 +72,7 @@ class MessageHandler(BaseHandler): depth = event.depth with (yield self.pagination_lock.write(room_id)): - yield self.store.purge_history(room_id, depth) + yield self.store.purge_history(room_id, depth, delete_local_events) @defer.inlineCallbacks def get_messages(self, requester, room_id=None, pagin_config=None, diff --git a/synapse/http/servlet.py b/synapse/http/servlet.py index 71420e54db..ef8e62901b 100644 --- a/synapse/http/servlet.py +++ b/synapse/http/servlet.py @@ -148,11 +148,13 @@ def parse_string_from_args(args, name, default=None, required=False, return default -def parse_json_value_from_request(request): +def parse_json_value_from_request(request, allow_empty_body=False): """Parse a JSON value from the body of a twisted HTTP request. Args: request: the twisted HTTP request. + allow_empty_body (bool): if True, an empty body will be accepted and + turned into None Returns: The JSON value. @@ -165,6 +167,9 @@ def parse_json_value_from_request(request): except Exception: raise SynapseError(400, "Error reading JSON content.") + if not content_bytes and allow_empty_body: + return None + try: content = simplejson.loads(content_bytes) except Exception as e: @@ -174,17 +179,24 @@ def parse_json_value_from_request(request): return content -def parse_json_object_from_request(request): +def parse_json_object_from_request(request, allow_empty_body=False): """Parse a JSON object from the body of a twisted HTTP request. Args: request: the twisted HTTP request. + allow_empty_body (bool): if True, an empty body will be accepted and + turned into an empty dict. Raises: SynapseError if the request body couldn't be decoded as JSON or if it wasn't a JSON object. """ - content = parse_json_value_from_request(request) + content = parse_json_value_from_request( + request, allow_empty_body=allow_empty_body, + ) + + if allow_empty_body and content is None: + return {} if type(content) != dict: message = "Content must be a JSON object." diff --git a/synapse/rest/client/v1/admin.py b/synapse/rest/client/v1/admin.py index 5022808ea9..f954d2ea65 100644 --- a/synapse/rest/client/v1/admin.py +++ b/synapse/rest/client/v1/admin.py @@ -128,7 +128,16 @@ class PurgeHistoryRestServlet(ClientV1RestServlet): if not is_admin: raise AuthError(403, "You are not a server admin") - yield self.handlers.message_handler.purge_history(room_id, event_id) + body = parse_json_object_from_request(request, allow_empty_body=True) + + delete_local_events = bool( + body.get("delete_local_history", False) + ) + + yield self.handlers.message_handler.purge_history( + room_id, event_id, + delete_local_events=delete_local_events, + ) defer.returnValue((200, {})) diff --git a/synapse/storage/events.py b/synapse/storage/events.py index 24d9978304..11a2ff2d8a 100644 --- a/synapse/storage/events.py +++ b/synapse/storage/events.py @@ -2031,16 +2031,32 @@ class EventsStore(SQLBaseStore): ) return self.runInteraction("get_all_new_events", get_all_new_events_txn) - def purge_history(self, room_id, topological_ordering): + def purge_history( + self, room_id, topological_ordering, delete_local_events, + ): """Deletes room history before a certain point + + Args: + room_id (str): + + topological_ordering (int): + minimum topo ordering to preserve + + delete_local_events (bool): + if True, we will delete local events as well as remote ones + (instead of just marking them as outliers and deleting their + state groups). """ return self.runInteraction( "purge_history", - self._purge_history_txn, room_id, topological_ordering + self._purge_history_txn, room_id, topological_ordering, + delete_local_events, ) - def _purge_history_txn(self, txn, room_id, topological_ordering): + def _purge_history_txn( + self, txn, room_id, topological_ordering, delete_local_events, + ): # Tables that should be pruned: # event_auth # event_backward_extremities @@ -2093,11 +2109,14 @@ class EventsStore(SQLBaseStore): to_delete = [ (event_id,) for event_id, state_key in event_rows - if state_key is None and not self.hs.is_mine_id(event_id) + if state_key is None and ( + delete_local_events or not self.hs.is_mine_id(event_id) + ) ] logger.info( - "[purge] found %i events before cutoff, of which %i are remote" - " non-state events to delete", len(event_rows), len(to_delete)) + "[purge] found %i events before cutoff, of which %i can be deleted", + len(event_rows), len(to_delete), + ) logger.info("[purge] Finding new backward extremities") @@ -2273,7 +2292,9 @@ class EventsStore(SQLBaseStore): " WHERE event_id = ?", [ (True, event_id,) for event_id, state_key in event_rows - if state_key is not None or self.hs.is_mine_id(event_id) + if state_key is not None or ( + not delete_local_events and self.hs.is_mine_id(event_id) + ) ] ) -- cgit 1.5.1 From dbe80a286b85f9427763344c260921745c2ca78d Mon Sep 17 00:00:00 2001 From: Richard van der Hoff Date: Fri, 9 Mar 2018 16:17:27 +0000 Subject: refactor JsonResource rephrase the OPTIONS and unrecognised request handling so that they look similar to the common flow. --- synapse/http/server.py | 78 +++++++++++++++++++++++++++++--------------------- 1 file changed, 46 insertions(+), 32 deletions(-) (limited to 'synapse/http') diff --git a/synapse/http/server.py b/synapse/http/server.py index 165c684d0d..d774476e5b 100644 --- a/synapse/http/server.py +++ b/synapse/http/server.py @@ -1,5 +1,6 @@ # -*- coding: utf-8 -*- # Copyright 2014-2016 OpenMarket Ltd +# 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. @@ -276,49 +277,54 @@ class JsonResource(HttpServer, resource.Resource): This checks if anyone has registered a callback for that method and path. """ - if request.method == "OPTIONS": - self._send_response(request, 200, {}) - return + callback, group_dict = self._get_handler_for_request(request) - # Loop through all the registered callbacks to check if the method - # and path regex match - for path_entry in self.path_regexs.get(request.method, []): - m = path_entry.pattern.match(request.path) - if not m: - continue + servlet_instance = getattr(callback, "__self__", None) + if servlet_instance is not None: + servlet_classname = servlet_instance.__class__.__name__ + else: + servlet_classname = "%r" % callback - # We found a match! First update the metrics object to indicate - # which servlet is handling the request. + request_metrics.name = servlet_classname - callback = path_entry.callback + # Now trigger the callback. If it returns a response, we send it + # here. If it throws an exception, that is handled by the wrapper + # installed by @request_handler. - servlet_instance = getattr(callback, "__self__", None) - if servlet_instance is not None: - servlet_classname = servlet_instance.__class__.__name__ - else: - servlet_classname = "%r" % callback + kwargs = intern_dict({ + name: urllib.unquote(value).decode("UTF-8") if value else value + for name, value in group_dict.items() + }) - request_metrics.name = servlet_classname + callback_return = yield callback(request, **kwargs) + if callback_return is not None: + code, response = callback_return + self._send_response(request, code, response) - # Now trigger the callback. If it returns a response, we send it - # here. If it throws an exception, that is handled by the wrapper - # installed by @request_handler. + def _get_handler_for_request(self, request): + """Finds a callback method to handle the given request - kwargs = intern_dict({ - name: urllib.unquote(value).decode("UTF-8") if value else value - for name, value in m.groupdict().items() - }) + Args: + request (twisted.web.http.Request): - callback_return = yield callback(request, **kwargs) - if callback_return is not None: - code, response = callback_return - self._send_response(request, code, response) + Returns: + Tuple[Callable, dict[str, str]]: callback method, and the dict + mapping keys to path components as specified in the handler's + path match regexp + """ + if request.method == "OPTIONS": + return _options_handler, {} - return + # Loop through all the registered callbacks to check if the method + # and path regex match + for path_entry in self.path_regexs.get(request.method, []): + m = path_entry.pattern.match(request.path) + if m: + # We found a match! + return path_entry.callback, m.groupdict() # Huh. No one wanted to handle that? Fiiiiiine. Send 400. - request_metrics.name = self.__class__.__name__ + ".UnrecognizedRequest" - raise UnrecognizedRequestError() + return _unrecognised_request_handler, {} def _send_response(self, request, code, response_json_object, response_code_message=None): @@ -335,6 +341,14 @@ class JsonResource(HttpServer, resource.Resource): ) +def _options_handler(request): + return {} + + +def _unrecognised_request_handler(request): + raise UnrecognizedRequestError() + + class RequestMetrics(object): def start(self, clock, name): self.start = clock.time_msec() -- cgit 1.5.1 From 88541f9009a7ca39c85cac7483d6a240ef497d33 Mon Sep 17 00:00:00 2001 From: Richard van der Hoff Date: Fri, 9 Mar 2018 16:19:18 +0000 Subject: Add a metric which increments when a request is received It's useful to know when there are peaks in incoming requests - which isn't quite the same as there being peaks in outgoing responses, due to the time taken to handle requests. --- synapse/http/server.py | 12 ++++++++++-- synapse/metrics/__init__.py | 16 ++++++++++++++++ 2 files changed, 26 insertions(+), 2 deletions(-) (limited to 'synapse/http') diff --git a/synapse/http/server.py b/synapse/http/server.py index d774476e5b..6c5d8bb556 100644 --- a/synapse/http/server.py +++ b/synapse/http/server.py @@ -60,6 +60,11 @@ response_count = metrics.register_counter( ) ) +requests_counter = metrics.register_counter( + "requests_received", + labels=["method", "servlet", ], +) + outgoing_responses_counter = metrics.register_counter( "responses", labels=["method", "code"], @@ -146,7 +151,8 @@ def wrap_request_handler(request_handler, include_metrics=False): # at the servlet name. For most requests that name will be # JsonResource (or a subclass), and JsonResource._async_render # will update it once it picks a servlet. - request_metrics.start(self.clock, name=self.__class__.__name__) + servlet_name = self.__class__.__name__ + request_metrics.start(self.clock, name=servlet_name) request_context.request = request_id with request.processing(): @@ -155,6 +161,7 @@ def wrap_request_handler(request_handler, include_metrics=False): if include_metrics: yield request_handler(self, request, request_metrics) else: + requests_counter.inc(request.method, servlet_name) yield request_handler(self, request) except CodeMessageException as e: code = e.code @@ -286,6 +293,7 @@ class JsonResource(HttpServer, resource.Resource): servlet_classname = "%r" % callback request_metrics.name = servlet_classname + requests_counter.inc(request.method, servlet_classname) # Now trigger the callback. If it returns a response, we send it # here. If it throws an exception, that is handled by the wrapper @@ -342,7 +350,7 @@ class JsonResource(HttpServer, resource.Resource): def _options_handler(request): - return {} + return 200, {} def _unrecognised_request_handler(request): diff --git a/synapse/metrics/__init__.py b/synapse/metrics/__init__.py index e0cfb7d08f..50d99d7a5c 100644 --- a/synapse/metrics/__init__.py +++ b/synapse/metrics/__init__.py @@ -57,15 +57,31 @@ class Metrics(object): return metric def register_counter(self, *args, **kwargs): + """ + Returns: + CounterMetric + """ return self._register(CounterMetric, *args, **kwargs) def register_callback(self, *args, **kwargs): + """ + Returns: + CallbackMetric + """ return self._register(CallbackMetric, *args, **kwargs) def register_distribution(self, *args, **kwargs): + """ + Returns: + DistributionMetric + """ return self._register(DistributionMetric, *args, **kwargs) def register_cache(self, *args, **kwargs): + """ + Returns: + CacheMetric + """ return self._register(CacheMetric, *args, **kwargs) -- cgit 1.5.1 From 58dd148c4f67e1cb6b150bf43a437b33089cfe5e Mon Sep 17 00:00:00 2001 From: Richard van der Hoff Date: Fri, 9 Mar 2018 18:05:41 +0000 Subject: Add some docstrings to help figure this out --- synapse/http/server.py | 28 ++++++++++++++++++++++++++-- 1 file changed, 26 insertions(+), 2 deletions(-) (limited to 'synapse/http') diff --git a/synapse/http/server.py b/synapse/http/server.py index 6c5d8bb556..4b567215c8 100644 --- a/synapse/http/server.py +++ b/synapse/http/server.py @@ -237,7 +237,7 @@ class JsonResource(HttpServer, resource.Resource): """ This implements the HttpServer interface and provides JSON support for Resources. - Register callbacks via register_path() + Register callbacks via register_paths() Callbacks can return a tuple of status code and a dict in which case the the dict will automatically be sent to the client as a JSON object. @@ -318,7 +318,11 @@ class JsonResource(HttpServer, resource.Resource): Returns: Tuple[Callable, dict[str, str]]: callback method, and the dict mapping keys to path components as specified in the handler's - path match regexp + path match regexp. + + The callback will normally be a method registered via + register_paths, so will return (possibly via Deferred) either + None, or a tuple of (http code, response body). """ if request.method == "OPTIONS": return _options_handler, {} @@ -350,10 +354,30 @@ class JsonResource(HttpServer, resource.Resource): def _options_handler(request): + """Request handler for OPTIONS requests + + This is a request handler suitable for return from + _get_handler_for_request. It returns a 200 and an empty body. + + Args: + request (twisted.web.http.Request): + + Returns: + Tuple[int, dict]: http code, response body. + """ return 200, {} def _unrecognised_request_handler(request): + """Request handler for unrecognised requests + + This is a request handler suitable for return from + _get_handler_for_request. It actually just raises an + UnrecognizedRequestError. + + Args: + request (twisted.web.http.Request): + """ raise UnrecognizedRequestError() -- cgit 1.5.1 From a8ce159be43560e9aea8f3be65110eea49d1f50e Mon Sep 17 00:00:00 2001 From: Erik Johnston Date: Thu, 15 Mar 2018 23:38:43 +0000 Subject: Replace some ujson with simplejson to make it work --- synapse/http/server.py | 3 ++- synapse/rest/client/v2_alpha/sync.py | 2 +- synapse/storage/events.py | 2 +- synapse/storage/events_worker.py | 2 +- 4 files changed, 5 insertions(+), 4 deletions(-) (limited to 'synapse/http') diff --git a/synapse/http/server.py b/synapse/http/server.py index 4b567215c8..3c7a0ef97a 100644 --- a/synapse/http/server.py +++ b/synapse/http/server.py @@ -38,6 +38,7 @@ import collections import logging import urllib import ujson +import simplejson logger = logging.getLogger(__name__) @@ -462,7 +463,7 @@ def respond_with_json(request, code, json_object, send_cors=False, json_bytes = encode_canonical_json(json_object) else: # ujson doesn't like frozen_dicts. - json_bytes = ujson.dumps(json_object, ensure_ascii=False) + json_bytes = simplejson.dumps(json_object) return respond_with_json_bytes( request, code, json_bytes, diff --git a/synapse/rest/client/v2_alpha/sync.py b/synapse/rest/client/v2_alpha/sync.py index a0a8e4b8e4..eb91c0b293 100644 --- a/synapse/rest/client/v2_alpha/sync.py +++ b/synapse/rest/client/v2_alpha/sync.py @@ -33,7 +33,7 @@ from ._base import set_timeline_upper_limit import itertools import logging -import ujson as json +import simplejson as json logger = logging.getLogger(__name__) diff --git a/synapse/storage/events.py b/synapse/storage/events.py index 3890878170..9fc65229fd 100644 --- a/synapse/storage/events.py +++ b/synapse/storage/events.py @@ -38,7 +38,7 @@ from functools import wraps import synapse.metrics import logging -import ujson as json +import simplejson as json # these are only included to make the type annotations work from synapse.events import EventBase # noqa: F401 diff --git a/synapse/storage/events_worker.py b/synapse/storage/events_worker.py index 86c3b48ad4..2e23dd78ba 100644 --- a/synapse/storage/events_worker.py +++ b/synapse/storage/events_worker.py @@ -28,7 +28,7 @@ from synapse.api.errors import SynapseError from collections import namedtuple import logging -import ujson as json +import simplejson as json # these are only included to make the type annotations work from synapse.events import EventBase # noqa: F401 -- cgit 1.5.1