From a3e9bef68557c45a9823cb6203660713f05cae9e Mon Sep 17 00:00:00 2001
From: Brian Cloutier <github@briancloutier.com>
Date: Fri, 4 Aug 2017 17:37:49 +0300
Subject: [PATCH] All users of WorkerNodeHash take an AccessShareLock

The metadata cache simulates a SELECT on pg_dist_node. Now the locks it
takes also simulate that SELECT.
---
 src/backend/distributed/master/master_metadata_utility.c | 2 --
 src/backend/distributed/utils/metadata_cache.c           | 7 +++++++
 src/backend/distributed/utils/reference_table_utils.c    | 1 -
 3 files changed, 7 insertions(+), 3 deletions(-)

diff --git a/src/backend/distributed/master/master_metadata_utility.c b/src/backend/distributed/master/master_metadata_utility.c
index 084ee1555..f2bbab8b2 100644
--- a/src/backend/distributed/master/master_metadata_utility.c
+++ b/src/backend/distributed/master/master_metadata_utility.c
@@ -174,8 +174,6 @@ DistributedTableSize(Oid relationId, char *sizeQuery)
 
 	ErrorIfNotSuitableToGetSize(relationId);
 
-	LockRelationOid(DistNodeRelationId(), AccessShareLock);
-
 	workerNodeList = ActivePrimaryNodeList();
 
 	foreach(workerNodeCell, workerNodeList)
diff --git a/src/backend/distributed/utils/metadata_cache.c b/src/backend/distributed/utils/metadata_cache.c
index b0980bee3..b35606f1c 100644
--- a/src/backend/distributed/utils/metadata_cache.c
+++ b/src/backend/distributed/utils/metadata_cache.c
@@ -43,6 +43,7 @@
 #include "nodes/makefuncs.h"
 #include "parser/parse_func.h"
 #include "parser/parse_type.h"
+#include "storage/lmgr.h"
 #include "utils/builtins.h"
 #include "utils/catcache.h"
 #include "utils/datum.h"
@@ -2272,6 +2273,12 @@ GetWorkerNodeHash(void)
 {
 	InitializeCaches(); /* ensure relevant callbacks are registered */
 
+	/*
+	 * Simulate a SELECT from pg_dist_node, ensure pg_dist_node doesn't change while our
+	 * caller is using WorkerNodeHash.
+	 */
+	LockRelationOid(DistNodeRelationId(), AccessShareLock);
+
 	/*
 	 * We might have some concurrent metadata changes. In order to get the changes,
 	 * we first need to accept the cache invalidation messages.
diff --git a/src/backend/distributed/utils/reference_table_utils.c b/src/backend/distributed/utils/reference_table_utils.c
index a9172a0a1..d0362c88e 100644
--- a/src/backend/distributed/utils/reference_table_utils.c
+++ b/src/backend/distributed/utils/reference_table_utils.c
@@ -230,7 +230,6 @@ ReplicateShardToAllWorkers(ShardInterval *shardInterval)
 	List *workerNodeList = NULL;
 	ListCell *workerNodeCell = NULL;
 
-	LockRelationOid(DistNodeRelationId(), AccessShareLock);
 	workerNodeList = ActivePrimaryNodeList();
 
 	/*