From ec1a6b0bc6e9616c02acbf9e63a00746019bc83e Mon Sep 17 00:00:00 2001
From: Bharath Rupireddy <bharath.rupireddyforpostgres@gmail.com>
Date: Tue, 14 Dec 2021 13:31:26 +0000
Subject: [PATCH v1] Add LOG messages when replication slots become active and
 inactive

These logs will be extremely useful on production servers to debug
and analyze inactive replication slot issues.
---
 src/backend/replication/slot.c | 39 ++++++++++++++++++++++++++++++++++
 1 file changed, 39 insertions(+)

diff --git a/src/backend/replication/slot.c b/src/backend/replication/slot.c
index 90ba9b417d..63bf1140ec 100644
--- a/src/backend/replication/slot.c
+++ b/src/backend/replication/slot.c
@@ -318,6 +318,11 @@ ReplicationSlotCreate(const char *name, bool db_specific,
 	SpinLockRelease(&slot->mutex);
 	MyReplicationSlot = slot;
 
+	ereport(LOG,
+			(errmsg("replication slot \"%s\" becomes active",
+					NameStr(slot->data.name)),
+			 errdetail("The process with PID %d acquired it.", slot->active_pid)));
+
 	LWLockRelease(ReplicationSlotControlLock);
 
 	/*
@@ -457,6 +462,11 @@ retry:
 
 	/* We made this slot active, so it's ours now. */
 	MyReplicationSlot = s;
+
+	ereport(LOG,
+			(errmsg("replication slot \"%s\" becomes active",
+					NameStr(s->data.name)),
+			errdetail("The process with PID %d acquired it.", s->active_pid)));
 }
 
 /*
@@ -499,6 +509,10 @@ ReplicationSlotRelease(void)
 
 	if (slot->data.persistency == RS_PERSISTENT)
 	{
+		int	pid;
+
+		pid = slot->active_pid;
+
 		/*
 		 * Mark persistent slot inactive.  We're not freeing it, just
 		 * disconnecting, but wake up others that may be waiting for it.
@@ -506,6 +520,12 @@ ReplicationSlotRelease(void)
 		SpinLockAcquire(&slot->mutex);
 		slot->active_pid = 0;
 		SpinLockRelease(&slot->mutex);
+
+		ereport(LOG,
+				(errmsg("replication slot \"%s\" becomes inactive",
+						NameStr(slot->data.name)),
+				 errdetail("The process with PID %d released it.", pid)));
+
 		ConditionVariableBroadcast(&slot->active_cv);
 	}
 
@@ -594,6 +614,7 @@ ReplicationSlotDropPtr(ReplicationSlot *slot)
 {
 	char		path[MAXPGPATH];
 	char		tmppath[MAXPGPATH];
+	int			pid;
 
 	/*
 	 * If some other backend ran this code concurrently with us, we might try
@@ -606,6 +627,8 @@ ReplicationSlotDropPtr(ReplicationSlot *slot)
 	sprintf(path, "pg_replslot/%s", NameStr(slot->data.name));
 	sprintf(tmppath, "pg_replslot/%s.tmp", NameStr(slot->data.name));
 
+	pid = slot->active_pid;
+
 	/*
 	 * Rename the slot directory on disk, so that we'll no longer recognize
 	 * this as a valid slot.  Note that if this fails, we've got to mark the
@@ -636,6 +659,14 @@ ReplicationSlotDropPtr(ReplicationSlot *slot)
 		slot->active_pid = 0;
 		SpinLockRelease(&slot->mutex);
 
+		ereport(LOG,
+				(errmsg("replication slot \"%s\" becomes inactive",
+						NameStr(slot->data.name)),
+				 errdetail("The process with PID %d released it.", pid)));
+
+		/* let's not emit the above message twice */
+		pid = -1;
+
 		/* wake up anyone waiting on this slot */
 		ConditionVariableBroadcast(&slot->active_cv);
 
@@ -645,6 +676,7 @@ ReplicationSlotDropPtr(ReplicationSlot *slot)
 						path, tmppath)));
 	}
 
+
 	/*
 	 * The slot is definitely gone.  Lock out concurrent scans of the array
 	 * long enough to kill it.  It's OK to clear the active PID here without
@@ -658,6 +690,13 @@ ReplicationSlotDropPtr(ReplicationSlot *slot)
 	slot->active_pid = 0;
 	slot->in_use = false;
 	LWLockRelease(ReplicationSlotControlLock);
+
+	if (pid > 0)
+		ereport(LOG,
+				(errmsg("replication slot \"%s\" becomes inactive",
+						NameStr(slot->data.name)),
+				 errdetail("The process with PID %d released it.", pid)));
+
 	ConditionVariableBroadcast(&slot->active_cv);
 
 	/*
-- 
2.25.1