From 90e472054c46be9d42ea36aea5683d87b1f11420 Mon Sep 17 00:00:00 2001
From: Dennis Felsing <dennis@felsing.org>
Date: Wed, 15 Jan 2025 16:27:50 +0000
Subject: [PATCH] DNM: Try out jemalloc leak profiling

Motiviation is to find memory leaks in CI, want to evaluate if this
works at all first

See https://github.com/jemalloc/jemalloc/wiki/Use-Case%3A-Leak-Checking
---
 ci/plugins/mzcompose/hooks/command                | 2 ++
 ci/plugins/mzcompose/hooks/pre-exit               | 4 ++++
 misc/python/materialize/cli/ci_annotate_errors.py | 1 +
 src/prof/src/jemalloc.rs                          | 3 ++-
 4 files changed, 9 insertions(+), 1 deletion(-)

diff --git a/ci/plugins/mzcompose/hooks/command b/ci/plugins/mzcompose/hooks/command
index 7086c52327b92..7fbf19210e060 100644
--- a/ci/plugins/mzcompose/hooks/command
+++ b/ci/plugins/mzcompose/hooks/command
@@ -45,6 +45,8 @@ if [ -n "${CI_COVERAGE_ENABLED:-}" ]; then
   find . -name '*.profraw' -delete # Remove remaining profraw files from coverage runs
 fi
 
+find . -name 'jeprof*.heap' -delete
+
 ci_collapsed_heading ":docker: Purging all existing docker containers and volumes, regardless of origin"
 docker ps --all --quiet | xargs --no-run-if-empty docker rm --force --volumes
 rm -f services.log
diff --git a/ci/plugins/mzcompose/hooks/pre-exit b/ci/plugins/mzcompose/hooks/pre-exit
index 5153e85197d18..90a2173821169 100755
--- a/ci/plugins/mzcompose/hooks/pre-exit
+++ b/ci/plugins/mzcompose/hooks/pre-exit
@@ -132,6 +132,10 @@ if [ -n "${CI_COVERAGE_ENABLED:-}" ] && [ -z "${BUILDKITE_MZCOMPOSE_PLUGIN_SKIP_
     fi
 fi
 
+find . -name 'jeprof*.heap' | while read -r i; do
+    buildkite-agent artifact upload "$i"
+done
+
 if [ "$CI_ANNOTATE_ERRORS_RESULT" -ne 0 ]; then
   echo "+++ ci-annotate-errors failed, which indicates that an unknown error was found"
   exit "$CI_ANNOTATE_ERRORS_RESULT"
diff --git a/misc/python/materialize/cli/ci_annotate_errors.py b/misc/python/materialize/cli/ci_annotate_errors.py
index 53b7fc82d272b..0c1c404e16401 100644
--- a/misc/python/materialize/cli/ci_annotate_errors.py
+++ b/misc/python/materialize/cli/ci_annotate_errors.py
@@ -101,6 +101,7 @@
     | (FAIL|TIMEOUT)\s+\[\s*\d+\.\d+s\]
     # parallel-workload
     | worker_.*\ still\ running: [\s\S]* Threads\ have\ not\ stopped\ within\ 5\ minutes,\ exiting\ hard
+    | Leak\ approximation\ summary:
     )
     .* $
     """,
diff --git a/src/prof/src/jemalloc.rs b/src/prof/src/jemalloc.rs
index c02c035325113..4b33abe7eb6ee 100644
--- a/src/prof/src/jemalloc.rs
+++ b/src/prof/src/jemalloc.rs
@@ -26,7 +26,8 @@ use tikv_jemalloc_ctl::{epoch, stats};
 
 #[allow(non_upper_case_globals)]
 #[export_name = "malloc_conf"]
-pub static malloc_conf: &[u8] = b"prof:true,prof_active:true,lg_prof_sample:19\0";
+pub static malloc_conf: &[u8] =
+    b"prof_leak:true,prof_final:true,prof:true,prof_active:true,lg_prof_sample:19\0";
 
 #[derive(Copy, Clone, Debug)]
 pub struct JemallocProfMetadata {