From ca597206b8ebd52b29cec8ddc1e2591211fb22d6 Mon Sep 17 00:00:00 2001
From: Christian Schwarz <christian@neon.tech>
Date: Wed, 29 Nov 2023 19:44:37 +0100
Subject: [PATCH] walredo: latency histogram for spawn duration (#5925)

fixes https://github.com/neondatabase/neon/issues/5891
---
 pageserver/src/metrics.rs | 10 ++++++++++
 pageserver/src/walredo.rs |  6 +++++-
 2 files changed, 15 insertions(+), 1 deletion(-)

diff --git a/pageserver/src/metrics.rs b/pageserver/src/metrics.rs
index b178024b9a..c15b1fce63 100644
--- a/pageserver/src/metrics.rs
+++ b/pageserver/src/metrics.rs
@@ -1253,6 +1253,15 @@ pub(crate) static WAL_REDO_RECORD_COUNTER: Lazy<IntCounter> = Lazy::new(|| {
     .unwrap()
 });
 
+pub(crate) static WAL_REDO_PROCESS_LAUNCH_DURATION_HISTOGRAM: Lazy<Histogram> = Lazy::new(|| {
+    register_histogram!(
+        "pageserver_wal_redo_process_launch_duration",
+        "Histogram of the duration of successful WalRedoProcess::launch calls",
+        redo_histogram_time_buckets!(),
+    )
+    .expect("failed to define a metric")
+});
+
 pub(crate) struct WalRedoProcessCounters {
     pub(crate) started: IntCounter,
     pub(crate) killed_by_cause: enum_map::EnumMap<WalRedoKillCause, IntCounter>,
@@ -1962,6 +1971,7 @@ pub fn preinitialize_metrics() {
         &WAL_REDO_TIME,
         &WAL_REDO_RECORDS_HISTOGRAM,
         &WAL_REDO_BYTES_HISTOGRAM,
+        &WAL_REDO_PROCESS_LAUNCH_DURATION_HISTOGRAM,
     ]
     .into_iter()
     .for_each(|h| {
diff --git a/pageserver/src/walredo.rs b/pageserver/src/walredo.rs
index ed468f220e..edce158e75 100644
--- a/pageserver/src/walredo.rs
+++ b/pageserver/src/walredo.rs
@@ -47,7 +47,8 @@ use pageserver_api::shard::TenantShardId;
 use crate::config::PageServerConf;
 use crate::metrics::{
     WalRedoKillCause, WAL_REDO_BYTES_HISTOGRAM, WAL_REDO_PROCESS_COUNTERS,
-    WAL_REDO_RECORDS_HISTOGRAM, WAL_REDO_RECORD_COUNTER, WAL_REDO_TIME,
+    WAL_REDO_PROCESS_LAUNCH_DURATION_HISTOGRAM, WAL_REDO_RECORDS_HISTOGRAM,
+    WAL_REDO_RECORD_COUNTER, WAL_REDO_TIME,
 };
 use crate::pgdatadir_mapping::{key_to_rel_block, key_to_slru_block};
 use crate::repository::Key;
@@ -241,10 +242,13 @@ impl PostgresRedoManager {
                         let mut proc_guard = self.redo_process.write().unwrap();
                         match &*proc_guard {
                             None => {
+                                let timer =
+                                    WAL_REDO_PROCESS_LAUNCH_DURATION_HISTOGRAM.start_timer();
                                 let proc = Arc::new(
                                     WalRedoProcess::launch(self.conf, self.tenant_id, pg_version)
                                         .context("launch walredo process")?,
                                 );
+                                timer.observe_duration();
                                 *proc_guard = Some(Arc::clone(&proc));
                                 proc
                             }