Jason/otel exp

pyproject.toml

@@ -238,6 +238,7 @@ pydantic = [
     { version = "~=2.8", python = "<3.14" },
     { version = "~=2.12", python = ">=3.14" },
 ]
+opentelemetry-exporter-prometheus = "0.55b1"
 
 # This is for building the rust components during "poetry install", which
 # currently ignores the `build-system.requires` directive (c.f.

synapse/api/auth/__init__.py

@@ -20,13 +20,11 @@
 #
 from typing import TYPE_CHECKING, Optional, Protocol
 
-from prometheus_client import Histogram
-
 from twisted.web.server import Request
 
 from synapse.appservice import ApplicationService
 from synapse.http.site import SynapseRequest
-from synapse.metrics import SERVER_NAME_LABEL
+from synapse.metrics import SERVER_NAME_LABEL, SynapseHistogram
 from synapse.types import Requester
 
 if TYPE_CHECKING:
@@ -36,7 +34,7 @@
 GUEST_DEVICE_ID = "guest_device"
 
 
-introspection_response_timer = Histogram(
+introspection_response_timer = SynapseHistogram(
     "synapse_api_auth_delegated_introspection_response",
     "Time taken to get a response for an introspection request",
     labelnames=["code", SERVER_NAME_LABEL],

synapse/app/phone_stats_home.py

@@ -24,11 +24,9 @@
 import sys
 from typing import TYPE_CHECKING, Mapping, Sized
 
-from prometheus_client import Gauge
-
 from twisted.internet import defer
 
-from synapse.metrics import SERVER_NAME_LABEL
+from synapse.metrics import SERVER_NAME_LABEL, SynapseGauge
 from synapse.types import JsonDict
 from synapse.util.constants import (
     MILLISECONDS_PER_SECOND,
@@ -57,22 +55,22 @@
 _stats_process: list[tuple[int, "resource.struct_rusage"]] = []
 
 # Gauges to expose monthly active user control metrics
-current_mau_gauge = Gauge(
+current_mau_gauge = SynapseGauge(
     "synapse_admin_mau_current",
     "Current MAU",
     labelnames=[SERVER_NAME_LABEL],
 )
-current_mau_by_service_gauge = Gauge(
+current_mau_by_service_gauge = SynapseGauge(
     "synapse_admin_mau_current_mau_by_service",
     "Current MAU by service",
     labelnames=["app_service", SERVER_NAME_LABEL],
 )
-max_mau_gauge = Gauge(
+max_mau_gauge = SynapseGauge(
     "synapse_admin_mau_max",
     "MAU Limit",
     labelnames=[SERVER_NAME_LABEL],
 )
-registered_reserved_users_mau_gauge = Gauge(
+registered_reserved_users_mau_gauge = SynapseGauge(
     "synapse_admin_mau_registered_reserved_users",
     "Registered users with reserved threepids",
     labelnames=[SERVER_NAME_LABEL],

synapse/appservice/api.py

@@ -31,7 +31,6 @@
     Union,
 )
 
-from prometheus_client import Counter
 from typing_extensions import ParamSpec, TypeGuard
 
 from synapse.api.constants import EventTypes, Membership, ThirdPartyEntityKind
@@ -45,7 +44,7 @@
 from synapse.events.utils import SerializeEventConfig, serialize_event
 from synapse.http.client import SimpleHttpClient, is_unknown_endpoint
 from synapse.logging import opentracing
-from synapse.metrics import SERVER_NAME_LABEL
+from synapse.metrics import SERVER_NAME_LABEL, SynapseCounter
 from synapse.types import DeviceListUpdates, JsonDict, JsonMapping, ThirdPartyInstanceID
 from synapse.util.caches.response_cache import ResponseCache
 
@@ -54,31 +53,31 @@
 
 logger = logging.getLogger(__name__)
 
-sent_transactions_counter = Counter(
+sent_transactions_counter = SynapseCounter(
     "synapse_appservice_api_sent_transactions",
     "Number of /transactions/ requests sent",
     labelnames=["service", SERVER_NAME_LABEL],
 )
 
-failed_transactions_counter = Counter(
+failed_transactions_counter = SynapseCounter(
     "synapse_appservice_api_failed_transactions",
     "Number of /transactions/ requests that failed to send",
     labelnames=["service", SERVER_NAME_LABEL],
 )
 
-sent_events_counter = Counter(
+sent_events_counter = SynapseCounter(
     "synapse_appservice_api_sent_events",
     "Number of events sent to the AS",
     labelnames=["service", SERVER_NAME_LABEL],
 )
 
-sent_ephemeral_counter = Counter(
+sent_ephemeral_counter = SynapseCounter(
     "synapse_appservice_api_sent_ephemeral",
     "Number of ephemeral events sent to the AS",
     labelnames=["service", SERVER_NAME_LABEL],
 )
 
-sent_todevice_counter = Counter(
+sent_todevice_counter = SynapseCounter(
     "synapse_appservice_api_sent_todevice",
     "Number of todevice messages sent to the AS",
     labelnames=["service", SERVER_NAME_LABEL],

synapse/federation/federation_client.py

@@ -41,7 +41,6 @@
 )
 
 import attr
-from prometheus_client import Counter
 
 from synapse.api.constants import Direction, EventContentFields, EventTypes, Membership
 from synapse.api.errors import (
@@ -71,7 +70,7 @@
 from synapse.http.client import is_unknown_endpoint
 from synapse.http.types import QueryParams
 from synapse.logging.opentracing import SynapseTags, log_kv, set_tag, tag_args, trace
-from synapse.metrics import SERVER_NAME_LABEL
+from synapse.metrics import SERVER_NAME_LABEL, SynapseCounter
 from synapse.types import JsonDict, StrCollection, UserID, get_domain_from_id
 from synapse.types.handlers.policy_server import RECOMMENDATION_OK, RECOMMENDATION_SPAM
 from synapse.util.async_helpers import concurrently_execute
@@ -83,7 +82,7 @@
 
 logger = logging.getLogger(__name__)
 
-sent_queries_counter = Counter(
+sent_queries_counter = SynapseCounter(
     "synapse_federation_client_sent_queries", "", labelnames=["type", SERVER_NAME_LABEL]
 )
 

synapse/federation/federation_server.py

@@ -32,8 +32,6 @@
     Union,
 )
 
-from prometheus_client import Counter, Gauge, Histogram
-
 from twisted.python import failure
 
 from synapse.api.constants import (
@@ -79,7 +77,12 @@
     tag_args,
     trace,
 )
-from synapse.metrics import SERVER_NAME_LABEL
+from synapse.metrics import (
+    SERVER_NAME_LABEL,
+    SynapseCounter,
+    SynapseGauge,
+    SynapseHistogram,
+)
 from synapse.metrics.background_process_metrics import wrap_as_background_process
 from synapse.replication.http.federation import (
     ReplicationFederationSendEduRestServlet,
@@ -102,27 +105,27 @@
 
 logger = logging.getLogger(__name__)
 
-received_pdus_counter = Counter(
+received_pdus_counter = SynapseCounter(
     "synapse_federation_server_received_pdus", "", labelnames=[SERVER_NAME_LABEL]
 )
 
-received_edus_counter = Counter(
+received_edus_counter = SynapseCounter(
     "synapse_federation_server_received_edus", "", labelnames=[SERVER_NAME_LABEL]
 )
 
-received_queries_counter = Counter(
+received_queries_counter = SynapseCounter(
     "synapse_federation_server_received_queries",
     "",
     labelnames=["type", SERVER_NAME_LABEL],
 )
 
-pdu_process_time = Histogram(
+pdu_process_time = SynapseHistogram(
     "synapse_federation_server_pdu_process_time",
     "Time taken to process an event",
     labelnames=[SERVER_NAME_LABEL],
 )
 
-last_pdu_ts_metric = Gauge(
+last_pdu_ts_metric = SynapseGauge(
     "synapse_federation_last_received_pdu_time",
     "The timestamp of the last PDU which was successfully received from the given domain",
     labelnames=("origin_server_name", SERVER_NAME_LABEL),

synapse/federation/sender/__init__.py

@@ -142,7 +142,6 @@
 )
 
 import attr
-from prometheus_client import Counter
 
 from twisted.internet import defer
 
@@ -160,6 +159,7 @@
 from synapse.metrics import (
     SERVER_NAME_LABEL,
     LaterGauge,
+    SynapseCounter,
     event_processing_loop_counter,
     event_processing_loop_room_count,
     events_processed_counter,
@@ -184,13 +184,13 @@
 
 logger = logging.getLogger(__name__)
 
-sent_pdus_destination_dist_count = Counter(
+sent_pdus_destination_dist_count = SynapseCounter(
     "synapse_federation_client_sent_pdu_destinations_count",
     "Number of PDUs queued for sending to one or more destinations",
     labelnames=[SERVER_NAME_LABEL],
 )
 
-sent_pdus_destination_dist_total = Counter(
+sent_pdus_destination_dist_total = SynapseCounter(
     "synapse_federation_client_sent_pdu_destinations",
     "Total number of PDUs queued for sending across all destinations",
     labelnames=[SERVER_NAME_LABEL],

synapse/federation/sender/per_destination_queue.py

@@ -26,7 +26,6 @@
 from typing import TYPE_CHECKING, Hashable, Iterable, Optional
 
 import attr
-from prometheus_client import Counter
 
 from twisted.internet import defer
 
@@ -42,7 +41,7 @@
 from synapse.handlers.presence import format_user_presence_state
 from synapse.logging import issue9533_logger
 from synapse.logging.opentracing import SynapseTags, set_tag
-from synapse.metrics import SERVER_NAME_LABEL, sent_transactions_counter
+from synapse.metrics import SERVER_NAME_LABEL, SynapseCounter, sent_transactions_counter
 from synapse.types import JsonDict, ReadReceipt
 from synapse.util.retryutils import NotRetryingDestination, get_retry_limiter
 from synapse.visibility import filter_events_for_server
@@ -56,13 +55,13 @@
 logger = logging.getLogger(__name__)
 
 
-sent_edus_counter = Counter(
+sent_edus_counter = SynapseCounter(
     "synapse_federation_client_sent_edus",
     "Total number of EDUs successfully sent",
     labelnames=[SERVER_NAME_LABEL],
 )
 
-sent_edus_by_type = Counter(
+sent_edus_by_type = SynapseCounter(
     "synapse_federation_client_sent_edus_by_type",
     "Number of sent EDUs successfully sent, by event type",
     labelnames=["type", SERVER_NAME_LABEL],

synapse/federation/sender/transaction_manager.py

@@ -20,8 +20,6 @@
 import logging
 from typing import TYPE_CHECKING
 
-from prometheus_client import Gauge
-
 from synapse.api.constants import EduTypes
 from synapse.api.errors import HttpResponseException
 from synapse.events import EventBase
@@ -34,7 +32,7 @@
     tags,
     whitelisted_homeserver,
 )
-from synapse.metrics import SERVER_NAME_LABEL
+from synapse.metrics import SERVER_NAME_LABEL, SynapseGauge
 from synapse.types import JsonDict
 from synapse.util.json import json_decoder
 from synapse.util.metrics import measure_func
@@ -45,7 +43,7 @@
 logger = logging.getLogger(__name__)
 issue_8631_logger = logging.getLogger("synapse.8631_debug")
 
-last_pdu_ts_metric = Gauge(
+last_pdu_ts_metric = SynapseGauge(
     "synapse_federation_last_sent_pdu_time",
     "The timestamp of the last PDU which was successfully sent to the given domain",
     labelnames=("destination_server_name", SERVER_NAME_LABEL),