From 891408892ca09f994ad51b2a741f39305bd92ccb Mon Sep 17 00:00:00 2001
From: Cuong Nguyen <ctring23@gmail.com>
Date: Tue, 31 Oct 2023 18:23:00 -0400
Subject: [PATCH] Add metric for commit lock wait time

---
 pageserver/src/metrics.rs           | 11 +++++++++++
 pageserver/src/pgdatadir_mapping.rs |  4 ++++
 2 files changed, 15 insertions(+)

diff --git a/pageserver/src/metrics.rs b/pageserver/src/metrics.rs
index 4f18f4e622fc..8cdd6038ea8d 100644
--- a/pageserver/src/metrics.rs
+++ b/pageserver/src/metrics.rs
@@ -633,6 +633,8 @@ pub static LIVE_CONNECTIONS_COUNT: Lazy<IntGaugeVec> = Lazy::new(|| {
     .expect("failed to define a metric")
 });
 
+// Remotexact
+
 pub static LAST_RECEIVED_LSN: Lazy<IntGaugeVec> = Lazy::new(|| {
     register_int_gauge_vec!(
         "pageserver_last_received_lsn",
@@ -652,6 +654,15 @@ pub static LSN_RECEIVE_DELAY: Lazy<HistogramVec> = Lazy::new(|| {
     .expect("failed to define a metric")
 });
 
+pub static WAL_COMMIT_WRITER_LOCK_WAIT_TIME: Lazy<Histogram> = Lazy::new(|| {
+    register_histogram!(
+        "pageserver_wal_commit_writer_lock_wait_seconds",
+        "Time spent waiting for the commit writer lock",
+        CRITICAL_OP_BUCKETS.into(),
+    )
+    .expect("failed to define a metric")
+});
+
 // remote storage metrics
 
 /// NB: increment _after_ recording the current value into [`REMOTE_TIMELINE_CLIENT_CALLS_STARTED_HIST`].
diff --git a/pageserver/src/pgdatadir_mapping.rs b/pageserver/src/pgdatadir_mapping.rs
index d43c919464f1..4b72ed5fba59 100644
--- a/pageserver/src/pgdatadir_mapping.rs
+++ b/pageserver/src/pgdatadir_mapping.rs
@@ -9,6 +9,7 @@
 use super::tenant::{PageReconstructError, Timeline};
 use crate::context::RequestContext;
 use crate::keyspace::{KeySpace, KeySpaceAccum};
+use crate::metrics::WAL_COMMIT_WRITER_LOCK_WAIT_TIME;
 use crate::repository::*;
 use crate::walrecord::NeonWalRecord;
 use anyhow::Context;
@@ -1180,7 +1181,10 @@ impl<'a> DatadirModification<'a> {
     /// All the modifications in this atomic update are stamped by the specified LSN.
     ///
     pub async fn commit(&mut self) -> anyhow::Result<()> {
+        let timer = WAL_COMMIT_WRITER_LOCK_WAIT_TIME.start_timer();
         let writer = self.tline.writer().await;
+        timer.stop_and_record();
+
         let lsn = self.lsn;
         let pending_nblocks = self.pending_nblocks;
         self.pending_nblocks = 0;