Fix Prometheus recording rules to not use legacy metric names. (#13718)

changelog.d/13718.misc

@@ -0,0 +1 @@
+Add experimental configuration option to allow disabling legacy Prometheus metric names.

contrib/prometheus/synapse-v1.rules

@@ -1,21 +0,0 @@
-synapse_federation_transaction_queue_pendingEdus:total = sum(synapse_federation_transaction_queue_pendingEdus or absent(synapse_federation_transaction_queue_pendingEdus)*0)
-synapse_federation_transaction_queue_pendingPdus:total = sum(synapse_federation_transaction_queue_pendingPdus or absent(synapse_federation_transaction_queue_pendingPdus)*0)
-
-synapse_http_server_request_count:method{servlet=""} = sum(synapse_http_server_request_count) by (method)
-synapse_http_server_request_count:servlet{method=""} = sum(synapse_http_server_request_count) by (servlet)
-
-synapse_http_server_request_count:total{servlet=""} = sum(synapse_http_server_request_count:by_method) by (servlet)
-
-synapse_cache:hit_ratio_5m = rate(synapse_util_caches_cache:hits[5m]) / rate(synapse_util_caches_cache:total[5m])
-synapse_cache:hit_ratio_30s = rate(synapse_util_caches_cache:hits[30s]) / rate(synapse_util_caches_cache:total[30s])
-
-synapse_federation_client_sent{type="EDU"} = synapse_federation_client_sent_edus + 0
-synapse_federation_client_sent{type="PDU"} = synapse_federation_client_sent_pdu_destinations:count + 0
-synapse_federation_client_sent{type="Query"} = sum(synapse_federation_client_sent_queries) by (job)
-
-synapse_federation_server_received{type="EDU"} = synapse_federation_server_received_edus + 0
-synapse_federation_server_received{type="PDU"} = synapse_federation_server_received_pdus + 0
-synapse_federation_server_received{type="Query"} = sum(synapse_federation_server_received_queries) by (job)
-
-synapse_federation_transaction_queue_pending{type="EDU"} = synapse_federation_transaction_queue_pending_edus + 0
-synapse_federation_transaction_queue_pending{type="PDU"} = synapse_federation_transaction_queue_pending_pdus + 0

contrib/prometheus/synapse-v2.rules

@@ -20,18 +20,18 @@ groups:
     expr: 'sum(synapse_http_server_request_count:by_method) by (servlet)'
 
   - record: 'synapse_cache:hit_ratio_5m'
-    expr: 'rate(synapse_util_caches_cache:hits[5m]) / rate(synapse_util_caches_cache:total[5m])'
+    expr: 'rate(synapse_util_caches_cache_hits[5m]) / rate(synapse_util_caches_cache[5m])'
   - record: 'synapse_cache:hit_ratio_30s'
-    expr: 'rate(synapse_util_caches_cache:hits[30s]) / rate(synapse_util_caches_cache:total[30s])'
+    expr: 'rate(synapse_util_caches_cache_hits[30s]) / rate(synapse_util_caches_cache[30s])'
 
   - record: 'synapse_federation_client_sent'
     labels:
       type: "EDU"
-    expr: 'synapse_federation_client_sent_edus + 0'
+    expr: 'synapse_federation_client_sent_edus_total + 0'
   - record: 'synapse_federation_client_sent'
     labels:
       type: "PDU"
-    expr: 'synapse_federation_client_sent_pdu_destinations:count + 0'
+    expr: 'synapse_federation_client_sent_pdu_destinations_count_total + 0'
   - record: 'synapse_federation_client_sent'
     labels:
       type: "Query"
@@ -40,11 +40,11 @@ groups:
   - record: 'synapse_federation_server_received'
     labels:
       type: "EDU"
-    expr: 'synapse_federation_server_received_edus + 0'
+    expr: 'synapse_federation_server_received_edus_total + 0'
   - record: 'synapse_federation_server_received'
     labels:
       type: "PDU"
-    expr: 'synapse_federation_server_received_pdus + 0'
+    expr: 'synapse_federation_server_received_pdus_total + 0'
   - record: 'synapse_federation_server_received'
     labels:
       type: "Query"
@@ -60,19 +60,19 @@ groups:
     expr: 'synapse_federation_transaction_queue_pending_pdus + 0'
 
   - record: synapse_storage_events_persisted_by_source_type
-    expr: sum without(type, origin_type, origin_entity) (synapse_storage_events_persisted_events_sep{origin_type="remote"})
+    expr: sum without(type, origin_type, origin_entity) (synapse_storage_events_persisted_events_sep_total{origin_type="remote"})
     labels:
       type: remote
   - record: synapse_storage_events_persisted_by_source_type
-    expr: sum without(type, origin_type, origin_entity) (synapse_storage_events_persisted_events_sep{origin_entity="*client*",origin_type="local"})
+    expr: sum without(type, origin_type, origin_entity) (synapse_storage_events_persisted_events_sep_total{origin_entity="*client*",origin_type="local"})
     labels:
       type: local
   - record: synapse_storage_events_persisted_by_source_type
-    expr: sum without(type, origin_type, origin_entity) (synapse_storage_events_persisted_events_sep{origin_entity!="*client*",origin_type="local"})
+    expr: sum without(type, origin_type, origin_entity) (synapse_storage_events_persisted_events_sep_total{origin_entity!="*client*",origin_type="local"})
     labels:
       type: bridges
   - record: synapse_storage_events_persisted_by_event_type
-    expr: sum without(origin_entity, origin_type) (synapse_storage_events_persisted_events_sep)
+    expr: sum without(origin_entity, origin_type) (synapse_storage_events_persisted_events_sep_total)
   - record: synapse_storage_events_persisted_by_origin
-    expr: sum without(type) (synapse_storage_events_persisted_events_sep)
+    expr: sum without(type) (synapse_storage_events_persisted_events_sep_total)
 

synapse/app/phone_stats_home.py

@@ -32,15 +32,15 @@ logger = logging.getLogger("synapse.app.homeserver")
 _stats_process: List[Tuple[int, "resource.struct_rusage"]] = []
 
 # Gauges to expose monthly active user control metrics
-current_mau_gauge = Gauge("synapse_admin_mau:current", "Current MAU")
+current_mau_gauge = Gauge("synapse_admin_mau_current", "Current MAU")
 current_mau_by_service_gauge = Gauge(
     "synapse_admin_mau_current_mau_by_service",
     "Current MAU by service",
     ["app_service"],
 )
-max_mau_gauge = Gauge("synapse_admin_mau:max", "MAU Limit")
+max_mau_gauge = Gauge("synapse_admin_mau_max", "MAU Limit")
 registered_reserved_users_mau_gauge = Gauge(
-    "synapse_admin_mau:registered_reserved_users",
+    "synapse_admin_mau_registered_reserved_users",
     "Registered users with reserved threepids",
 )
 

synapse/federation/sender/__init__.py

@@ -62,12 +62,12 @@ if TYPE_CHECKING:
 logger = logging.getLogger(__name__)
 
 sent_pdus_destination_dist_count = Counter(
-    "synapse_federation_client_sent_pdu_destinations:count",
+    "synapse_federation_client_sent_pdu_destinations_count",
     "Number of PDUs queued for sending to one or more destinations",
 )
 
 sent_pdus_destination_dist_total = Counter(
-    "synapse_federation_client_sent_pdu_destinations:total",
+    "synapse_federation_client_sent_pdu_destinations",
     "Total number of PDUs queued for sending across all destinations",
 )
 

synapse/metrics/_legacy_exposition.py

@@ -93,6 +93,11 @@ LEGACY_METRIC_NAMES = {
     "synapse_util_caches_response_cache_hits": "synapse_util_caches_response_cache:hits",
     "synapse_util_caches_response_cache_evicted_size": "synapse_util_caches_response_cache:evicted_size",
     "synapse_util_caches_response_cache": "synapse_util_caches_response_cache:total",
+    "synapse_federation_client_sent_pdu_destinations": "synapse_federation_client_sent_pdu_destinations:total",
+    "synapse_federation_client_sent_pdu_destinations_count": "synapse_federation_client_sent_pdu_destinations:count",
+    "synapse_admin_mau_current": "synapse_admin_mau:current",
+    "synapse_admin_mau_max": "synapse_admin_mau:max",
+    "synapse_admin_mau_registered_reserved_users": "synapse_admin_mau:registered_reserved_users",
 }