Clear metadata OID cache prior to deadlock detection

src/backend/distributed/utils/maintenanced.c

@@ -266,6 +266,15 @@ CitusMaintenanceDaemonMain(Datum main_arg)
 
 		CHECK_FOR_INTERRUPTS();
 
+		/*
+		 * XXX: We clear the metadata cache before every iteration because otherwise
+		 * it might contain stale OIDs. It appears that in some cases invalidation
+		 * messages for a DROP EXTENSION may arrive during deadlock detection and
+		 * this causes us to cache a stale pg_dist_node OID. We'd actually expect
+		 * all invalidations to arrive after obtaining a lock in LockCitusExtension.
+		 */
+		ClearMetadataOIDCache();
+
 		/*
 		 * Perform Work.  If a specific task needs to be called sooner than
 		 * timeout indicates, it's ok to lower it to that value.  Expensive

src/backend/distributed/utils/metadata_cache.c

@@ -2694,11 +2694,21 @@ InvalidateDistRelationCacheCallback(Datum argument, Oid relationId)
 	 */
 	if (relationId != InvalidOid && relationId == MetadataCache.distPartitionRelationId)
 	{
-		memset(&MetadataCache, 0, sizeof(MetadataCache));
+		ClearMetadataOIDCache();
 	}
 }
 
 
+/*
+ * ClearMetadataOIDCache resets all the cached OIDs and the extensionLoaded flag.
+ */
+void
+ClearMetadataOIDCache(void)
+{
+	memset(&MetadataCache, 0, sizeof(MetadataCache));
+}
+
+
 /*
  * DistTableOidList iterates over the pg_dist_partition table and returns
  * a list that consists of the logicalrelids.

src/include/distributed/metadata_cache.h

@@ -86,6 +86,7 @@ extern List * DistTableOidList(void);
 extern List * ShardPlacementList(uint64 shardId);
 extern void CitusInvalidateRelcacheByRelid(Oid relationId);
 extern void CitusInvalidateRelcacheByShardId(int64 shardId);
+extern void ClearMetadataOIDCache(void);
 
 extern bool CitusHasBeenLoaded(void);
 extern bool CheckCitusVersion(int elevel);