fixes

synapse/app/homeserver.py

@@ -34,6 +34,7 @@ from synapse.module_api import ModuleApi
 from synapse.http.additional_resource import AdditionalResource
 from synapse.http.server import RootRedirect
 from synapse.http.site import SynapseSite
+from synapse.metrics import RegistryProxy
 from synapse.metrics.resource import METRICS_PREFIX
 from synapse.python_dependencies import CONDITIONAL_REQUIREMENTS, \
     check_requirements
@@ -60,6 +61,8 @@ from twisted.web.resource import EncodingResourceWrapper, NoResource
 from twisted.web.server import GzipEncoderFactory
 from twisted.web.static import File
 
+from prometheus_client.twisted import MetricsResource
+
 logger = logging.getLogger("synapse.app.homeserver")
 
 
@@ -229,8 +232,7 @@ class SynapseHomeServer(HomeServer):
             resources[WEB_CLIENT_PREFIX] = build_resource_for_web_client(self)
 
         if name == "metrics" and self.get_config().enable_metrics:
-            from prometheus_client.twisted import MetricsResource
-            resources[METRICS_PREFIX] = MetricsResource()
+            resources[METRICS_PREFIX] = MetricsResource(RegistryProxy())
 
         if name == "replication":
             resources[REPLICATION_PREFIX] = ReplicationRestResource(self)

synapse/federation/transaction_queue.py

@@ -64,7 +64,7 @@ class TransactionQueue(object):
         # done
         self.pending_transactions = {}
 
-        LaterGauge("pending_destinations", "", [],
+        LaterGauge("synapse_federation_client_pending_destinations", "", [],
             lambda: len(self.pending_transactions),
         )
 
@@ -89,11 +89,11 @@ class TransactionQueue(object):
         self.pending_edus_keyed_by_dest = edus_keyed = {}
 
         LaterGauge(
-            "pending_pdus", "", [],
+            "synapse_federation_client_pending_pdus", "", [],
             lambda: sum(map(len, pdus.values())),
         )
         LaterGauge(
-            "pending_edus", "", [],
+            "synapse_federation_client_pending_edus", "", [],
             lambda: (
                 sum(map(len, edus.values()))
                 + sum(map(len, presence.values()))

synapse/metrics/__init__.py

@@ -29,12 +29,20 @@ from twisted.internet import reactor
 
 logger = logging.getLogger(__name__)
 
-
 running_on_pypy = platform.python_implementation() == 'PyPy'
 all_metrics = []
 all_collectors = []
 all_gauges = {}
 
+
+class RegistryProxy(object):
+
+    def collect(self):
+        for metric in REGISTRY.collect():
+            if not metric.name.startswith("__"):
+                yield metric
+
+
 @attr.s(hash=True)
 class LaterGauge(object):
 
@@ -45,7 +53,7 @@ class LaterGauge(object):
 
     def collect(self):
 
-        g = GaugeMetricFamily(self.name, self.desc, self.labels)
+        g = GaugeMetricFamily(self.name, self.desc, labels=self.labels)
 
         try:
             calls = self.caller()

synapse/notifier.py

@@ -194,14 +194,14 @@ class Notifier(object):
                 all_user_streams.add(x)
 
             return sum(stream.count_listeners() for stream in all_user_streams)
-        LaterGauge("listeners", "", [], count_listeners)
+        LaterGauge("synapse_notifier_listeners", "", [], count_listeners)
 
         LaterGauge(
-            "rooms", "", [],
+            "synapse_notifier_rooms", "", [],
             lambda: count(bool, self.room_to_user_streams.values()),
         )
         LaterGauge(
-            "users", "", [],
+            "synapse_notifier_users", "", [],
             lambda: len(self.user_to_user_stream),
         )
 

synapse/push/httppusher.py

@@ -28,9 +28,9 @@ from prometheus_client import Counter
 
 logger = logging.getLogger(__name__)
 
-http_push_processed_counter = Counter("http_pushes_processed", "")
+http_push_processed_counter = Counter("synapse_http_httppusher_http_pushes_processed", "")
 
-http_push_failed_counter = Counter("http_pushes_failed", "")
+http_push_failed_counter = Counter("synapse_http_httppusher_http_pushes_failed", "")
 
 
 class HttpPusher(object):

synapse/util/caches/__init__.py

@@ -13,7 +13,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from prometheus_client.core import GaugeMetricFamily, REGISTRY
+from prometheus_client.core import Gauge, REGISTRY, GaugeMetricFamily
 
 import os
 
@@ -22,10 +22,20 @@ CACHE_SIZE_FACTOR = float(os.environ.get("SYNAPSE_CACHE_FACTOR", 0.5))
 caches_by_name = {}
 collectors_by_name = {}
 
-def register_cache(name, cache_name, cache):
+cache_size = Gauge("synapse_util_caches_cache:size", "", ["name"])
+cache_hits = Gauge("synapse_util_caches_cache:hits", "", ["name"])
+cache_evicted = Gauge("synapse_util_caches_cache:evicted_size", "", ["name"])
+cache_total = Gauge("synapse_util_caches_cache:total", "", ["name"])
+
+response_cache_size = Gauge("synapse_util_caches_response_cache:size", "", ["name"])
+response_cache_hits = Gauge("synapse_util_caches_response_cache:hits", "", ["name"])
+response_cache_evicted = Gauge("synapse_util_caches_response_cache:evicted_size", "", ["name"])
+response_cache_total = Gauge("synapse_util_caches_response_cache:total", "", ["name"])
+
+def register_cache(cache_type, cache_name, cache):
 
     # Check if the metric is already registered. Unregister it, if so.
-    metric_name = "synapse_util_caches_%s:%s" % (name, cache_name,)
+    metric_name = "cache_%s_%s" % (cache_type, cache_name,)
     if metric_name in collectors_by_name.keys():
         REGISTRY.unregister(collectors_by_name[metric_name])
 
@@ -44,15 +54,22 @@ def register_cache(name, cache_name, cache):
         def inc_evictions(self, size=1):
             self.evicted_size += size
 
-        def collect(self):
-            cache_size = len(cache)
+        def describe(self):
+            return []
 
-            gm = GaugeMetricFamily(metric_name, "", labels=["size", "hits", "misses", "total"])
-            gm.add_metric(["size"], cache_size)
-            gm.add_metric(["hits"], self.hits)
-            gm.add_metric(["misses"], self.misses)
-            gm.add_metric(["total"], self.hits + self.misses)
-            yield gm
+        def collect(self):
+            if cache_type == "response_cache":
+                response_cache_size.labels(cache_name).set(len(cache))
+                response_cache_hits.labels(cache_name).set(self.hits)
+                response_cache_evicted.labels(cache_name).set(self.evicted_size)
+                response_cache_total.labels(cache_name).set(self.hits + self.misses)
+            else:
+                cache_size.labels(cache_name).set(len(cache))
+                cache_hits.labels(cache_name).set(self.hits)
+                cache_evicted.labels(cache_name).set(self.evicted_size)
+                cache_total.labels(cache_name).set(self.hits + self.misses)
+
+            yield GaugeMetricFamily("__unused", "")
 
     metric = CacheMetric()
     REGISTRY.register(metric)
@@ -60,6 +77,7 @@ def register_cache(name, cache_name, cache):
     collectors_by_name[metric_name] = metric
     return metric
 
+
 KNOWN_KEYS = {
     key: key for key in
     (

synapse/util/caches/descriptors.py

@@ -80,7 +80,7 @@ class Cache(object):
         self.name = name
         self.keylen = keylen
         self.thread = None
-        self.metrics = register_cache("descriptor", name, self.cache)
+        self.metrics = register_cache("cache", name, self.cache)
 
     def _on_evicted(self, evicted_count):
         self.metrics.inc_evictions(evicted_count)