pageserver_live_connections: track as counter pair (#8227)

Generally counter pairs are preferred over gauges.
In this case, I found myself asking what the typical rate of accepted
page_service connections on a pageserver is, and I couldn't answer it
with the gauge metric.

There are a few dashboards using this metric:

https://github.com/search?q=repo%3Aneondatabase%2Fgrafana-dashboard-export%20pageserver_live_connections&type=code

I'll convert them to use the new metric once this PR reaches prod.

refs #7427

pageserver/src/metrics.rs

@@ -1456,10 +1456,12 @@ impl<'a, 'c> BasebackupQueryTimeOngoingRecording<'a, 'c> {
     }
 }
 
-pub(crate) static LIVE_CONNECTIONS_COUNT: Lazy<IntGaugeVec> = Lazy::new(|| {
-    register_int_gauge_vec!(
-        "pageserver_live_connections",
-        "Number of live network connections",
+pub(crate) static LIVE_CONNECTIONS: Lazy<IntCounterPairVec> = Lazy::new(|| {
+    register_int_counter_pair_vec!(
+        "pageserver_live_connections_started",
+        "Number of network connections that we started handling",
+        "pageserver_live_connections_finished",
+        "Number of network connections that we finished handling",
         &["pageserver_connection_kind"]
     )
     .expect("failed to define a metric")

pageserver/src/page_service.rs

@@ -55,7 +55,7 @@ use crate::basebackup::BasebackupError;
 use crate::context::{DownloadBehavior, RequestContext};
 use crate::import_datadir::import_wal_from_tar;
 use crate::metrics;
-use crate::metrics::{ComputeCommandKind, COMPUTE_COMMANDS_COUNTERS, LIVE_CONNECTIONS_COUNT};
+use crate::metrics::{ComputeCommandKind, COMPUTE_COMMANDS_COUNTERS, LIVE_CONNECTIONS};
 use crate::pgdatadir_mapping::Version;
 use crate::span::debug_assert_current_span_has_tenant_and_timeline_id;
 use crate::span::debug_assert_current_span_has_tenant_and_timeline_id_no_shard_id;
@@ -215,14 +215,9 @@ async fn page_service_conn_main(
     auth_type: AuthType,
     connection_ctx: RequestContext,
 ) -> anyhow::Result<()> {
-    // Immediately increment the gauge, then create a job to decrement it on task exit.
-    // One of the pros of `defer!` is that this will *most probably*
-    // get called, even in presence of panics.
-    let gauge = LIVE_CONNECTIONS_COUNT.with_label_values(&["page_service"]);
-    gauge.inc();
-    scopeguard::defer! {
-        gauge.dec();
-    }
+    let _guard = LIVE_CONNECTIONS
+        .with_label_values(&["page_service"])
+        .guard();
 
     socket
         .set_nodelay(true)

pageserver/src/tenant/timeline/walreceiver/walreceiver_connection.rs

@@ -26,7 +26,7 @@ use tracing::{debug, error, info, trace, warn, Instrument};
 use super::TaskStateUpdate;
 use crate::{
     context::RequestContext,
-    metrics::{LIVE_CONNECTIONS_COUNT, WALRECEIVER_STARTED_CONNECTIONS, WAL_INGEST},
+    metrics::{LIVE_CONNECTIONS, WALRECEIVER_STARTED_CONNECTIONS, WAL_INGEST},
     task_mgr::TaskKind,
     task_mgr::WALRECEIVER_RUNTIME,
     tenant::{debug_assert_current_span_has_tenant_and_timeline_id, Timeline, WalReceiverInfo},
@@ -208,14 +208,9 @@ pub(super) async fn handle_walreceiver_connection(
         .instrument(tracing::info_span!("poller")),
     );
 
-    // Immediately increment the gauge, then create a job to decrement it on task exit.
-    // One of the pros of `defer!` is that this will *most probably*
-    // get called, even in presence of panics.
-    let gauge = LIVE_CONNECTIONS_COUNT.with_label_values(&["wal_receiver"]);
-    gauge.inc();
-    scopeguard::defer! {
-        gauge.dec();
-    }
+    let _guard = LIVE_CONNECTIONS
+        .with_label_values(&["wal_receiver"])
+        .guard();
 
     let identify = identify_system(&replication_client).await?;
     info!("{identify:?}");