Don't use an arbitrary bound in histograms

I hit this trying to figure out how an mostly idle server was
reporting that no connection was idle for more than 30s.

Author: espindola

src/conn/pool/metrics.rs

@@ -68,7 +68,7 @@ impl MetricsHistogram {
 #[cfg(feature = "hdrhistogram")]
 impl Default for MetricsHistogram {
     fn default() -> Self {
-        let hdr = hdrhistogram::Histogram::new_with_bounds(1, 30 * 1_000_000, 2).unwrap();
+        let hdr = hdrhistogram::Histogram::new(2).unwrap();
         Self(std::sync::Mutex::new(hdr))
     }
 }

src/conn/pool/mod.rs

@@ -359,7 +359,8 @@ impl Pool {
                     .connection_idle_duration
                     .lock()
                     .unwrap()
-                    .saturating_record(since.elapsed().as_micros() as u64);
+                    .record(since.elapsed().as_micros() as u64)
+                    .ok();
                 #[cfg(feature = "hdrhistogram")]
                 let metrics = self.metrics();
                 conn.inner.active_since = Instant::now();
@@ -371,9 +372,8 @@ impl Pool {
                             .check_duration
                             .lock()
                             .unwrap()
-                            .saturating_record(
-                                conn.inner.active_since.elapsed().as_micros() as u64
-                            );
+                            .record(conn.inner.active_since.elapsed().as_micros() as u64)
+                            .ok();
                         Ok(conn)
                     }
                     .boxed(),
@@ -412,9 +412,8 @@ impl Pool {
                             .connect_duration
                             .lock()
                             .unwrap()
-                            .saturating_record(
-                                conn.inner.active_since.elapsed().as_micros() as u64
-                            );
+                            .record(conn.inner.active_since.elapsed().as_micros() as u64)
+                            .ok();
                     }
                     conn
                 }
@@ -1244,6 +1243,31 @@ mod test {
         Ok(())
     }
 
+    #[cfg(feature = "hdrhistogram")]
+    #[tokio::test]
+    async fn metrics() -> super::Result<()> {
+        let pool = pool_with_one_connection();
+
+        let metrics = pool.metrics();
+        let conn = pool.get_conn().await.unwrap();
+        tokio::time::sleep(Duration::from_millis(100)).await;
+        drop(conn);
+        pool.get_conn().await.unwrap();
+
+        let metrics = serde_json::to_value(&*metrics).unwrap();
+        let connection_active_duration = metrics
+            .get("connection_active_duration")
+            .unwrap()
+            .get("max")
+            .unwrap()
+            .as_u64()
+            .unwrap();
+        // We slept for 100 miliseconds holding a conneciton.
+        assert!(connection_active_duration > 100_000);
+
+        Ok(())
+    }
+
     #[cfg(feature = "nightly")]
     mod bench {
         use futures_util::future::{FutureExt, TryFutureExt};

src/conn/pool/recycler.rs

@@ -86,7 +86,8 @@ impl Future for Recycler {
                         .connection_active_duration
                         .lock()
                         .unwrap()
-                        .saturating_record($conn.inner.active_since.elapsed().as_micros() as u64);
+                        .record($conn.inner.active_since.elapsed().as_micros() as u64)
+                        .ok();
                     exchange.available.push_back($conn.into());
                     $self
                         .inner