From e1935f42a1d1de4948fa6f689f2fe4cd83d15549 Mon Sep 17 00:00:00 2001
From: Arseny Sher <sher-ars@yandex.ru>
Date: Mon, 27 Nov 2023 15:40:01 +0300
Subject: [PATCH] Don't generate core dump when walproposer intentionally
 panics.

Walproposer sometimes intentionally PANICs when its term is defeated as the
basebackup is likely spoiled by that time. We don't want core dumped in this
case.
---
 pgxn/neon/neon_utils.c  | 25 +++++++++++++++++++++++++
 pgxn/neon/neon_utils.h  |  1 +
 pgxn/neon/walproposer.c | 15 ++++++++++++++-
 3 files changed, 40 insertions(+), 1 deletion(-)

diff --git a/pgxn/neon/neon_utils.c b/pgxn/neon/neon_utils.c
index 06faea7490..807d2decf6 100644
--- a/pgxn/neon/neon_utils.c
+++ b/pgxn/neon/neon_utils.c
@@ -1,3 +1,6 @@
+
+#include <sys/resource.h>
+
 #include "postgres.h"
 
 #include "access/timeline.h"
@@ -114,3 +117,25 @@ pq_sendint64_le(StringInfo buf, uint64 i)
 	memcpy(buf->data + buf->len, &i, sizeof(uint64));
 	buf->len += sizeof(uint64);
 }
+
+/*
+ * Disables core dump for the current process.
+ */
+void
+disable_core_dump()
+{
+	struct rlimit rlim;
+
+#ifdef WALPROPOSER_LIB			/* skip in simulation mode */
+	return;
+#endif
+
+	rlim.rlim_cur = 0;
+	rlim.rlim_max = 0;
+	if (setrlimit(RLIMIT_CORE, &rlim))
+	{
+		int			save_errno = errno;
+
+		fprintf(stderr, "WARNING: disable cores setrlimit failed: %s", strerror(save_errno));
+	}
+}
diff --git a/pgxn/neon/neon_utils.h b/pgxn/neon/neon_utils.h
index e3fafc8d0f..20745d8b26 100644
--- a/pgxn/neon/neon_utils.h
+++ b/pgxn/neon/neon_utils.h
@@ -8,5 +8,6 @@ uint32		pq_getmsgint32_le(StringInfo msg);
 uint64		pq_getmsgint64_le(StringInfo msg);
 void		pq_sendint32_le(StringInfo buf, uint32 i);
 void		pq_sendint64_le(StringInfo buf, uint64 i);
+extern void disable_core_dump();
 
 #endif							/* __NEON_UTILS_H__ */
diff --git a/pgxn/neon/walproposer.c b/pgxn/neon/walproposer.c
index 7d9dbfdb7f..fc3332612c 100644
--- a/pgxn/neon/walproposer.c
+++ b/pgxn/neon/walproposer.c
@@ -35,6 +35,8 @@
  *
  *-------------------------------------------------------------------------
  */
+#include <sys/resource.h>
+
 #include "postgres.h"
 #include "libpq/pqformat.h"
 #include "neon.h"
@@ -1069,6 +1071,12 @@ DetermineEpochStartLsn(WalProposer *wp)
 			if (!((dth->n_entries >= 1) && (dth->entries[dth->n_entries - 1].term ==
 											walprop_shared->mineLastElectedTerm)))
 			{
+				/*
+				 * Panic to restart PG as we need to retake basebackup.
+				 * However, don't dump core as this is kinda expected
+				 * scenario.
+				 */
+				disable_core_dump();
 				walprop_log(PANIC,
 							"collected propEpochStartLsn %X/%X, but basebackup LSN %X/%X",
 							LSN_FORMAT_ARGS(wp->propEpochStartLsn),
@@ -1445,7 +1453,12 @@ RecvAppendResponses(Safekeeper *sk)
 
 		if (sk->appendResponse.term > wp->propTerm)
 		{
-			/* Another compute with higher term is running. */
+			/*
+			 * Another compute with higher term is running. Panic to restart
+			 * PG as we likely need to retake basebackup. However, don't dump
+			 * core as this is kinda expected scenario.
+			 */
+			disable_core_dump();
 			walprop_log(PANIC, "WAL acceptor %s:%s with term " INT64_FORMAT " rejected our request, our term " INT64_FORMAT "",
 						sk->host, sk->port,
 						sk->appendResponse.term, wp->propTerm);