diff --git a/synapse/util/logcontext.py b/synapse/util/logcontext.py
index 48c9f6802d..94fa7cac98 100644
--- a/synapse/util/logcontext.py
+++ b/synapse/util/logcontext.py
@@ -52,13 +52,17 @@ except Exception:
class LoggingContext(object):
"""Additional context for log formatting. Contexts are scoped within a
"with" block.
+
Args:
name (str): Name for the context for debugging.
"""
__slots__ = [
- "previous_context", "name", "usage_start", "usage_end", "main_thread",
- "__dict__", "tag", "alive",
+ "previous_context", "name", "ru_stime", "ru_utime",
+ "db_txn_count", "db_txn_duration_ms", "db_sched_duration_ms",
+ "usage_start", "usage_end",
+ "main_thread", "alive",
+ "request", "tag",
]
thread_local = threading.local()
@@ -83,6 +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
@@ -94,9 +101,17 @@ 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
+
+ # 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()
+ self.request = None
self.tag = ""
self.alive = True
@@ -105,7 +120,11 @@ class LoggingContext(object):
@classmethod
def current_context(cls):
- """Get the current logging context from thread local storage"""
+ """Get the current logging context from thread local storage
+
+ Returns:
+ LoggingContext: the current logging context
+ """
return getattr(cls.thread_local, "current_context", cls.sentinel)
@classmethod
@@ -155,11 +174,13 @@ class LoggingContext(object):
self.alive = False
def copy_to(self, record):
- """Copy fields from this context to the record"""
- for key, value in self.__dict__.items():
- setattr(record, key, value)
+ """Copy logging fields from this context to a log record or
+ another LoggingContext
+ """
- record.ru_utime, record.ru_stime = self.get_resource_usage()
+ # 'request' is the only field we currently use in the logger, so that's
+ # all we need to copy
+ record.request = self.request
def start(self):
if threading.current_thread() is not self.main_thread:
@@ -194,7 +215,16 @@ 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
+
+ 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):
diff --git a/synapse/util/metrics.py b/synapse/util/metrics.py
index 4ea930d3e8..059bb7fedf 100644
--- a/synapse/util/metrics.py
+++ b/synapse/util/metrics.py
@@ -27,25 +27,62 @@ 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"]
+# 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=(
+ metrics.name_prefix + "_block_db_txn_count:total",
+ ),
+)
+
+# 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"],
)
@@ -64,7 +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", "created_context"
+ "ru_stime",
+ "db_txn_count", "db_txn_duration_ms", "db_sched_duration_ms",
+ "created_context",
]
def __init__(self, clock, name):
@@ -84,7 +123,8 @@ 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
+ 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:
@@ -114,7 +154,12 @@ 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
+ )
+ block_db_sched_duration.inc_by(
+ (context.db_sched_duration_ms - self.db_sched_duration_ms) / 1000.,
+ self.name
)
if self.created_context:
|
