Fix a bug where we will pick up random nodes for individual replicated tables,

thus reducing the chances of query getting fully shipped to the remote node.

We now remember all nodes that can satify a READ request for a replicated table
and then finally choose a node randomly if no preferred datanode is specified.
This will avoid non-deterministic selection of FQS query plans as well as allow
us to send some more queries to the remote node.

diff --git a/src/backend/optimizer/util/pgxcship.c b/src/backend/optimizer/util/pgxcship.c
index 6d64434c9e14b213e7f5b1d57ffe0c242210dee6..7453bdc952cc91787f427a6e6c58622e3a67a94b 100644 (file)
--- a/src/backend/optimizer/util/pgxcship.c
+++ b/src/backend/optimizer/util/pgxcship.c
@@ -406,7 +406,8 @@ pgxc_FQS_find_datanodes(Query *query)
                 * preferred node choose that one, otherwise choose the first one.
                 */
                if (IsLocatorReplicated(exec_nodes->baselocatortype) &&
-                       exec_nodes->accesstype == RELATION_ACCESS_READ)
+                       (exec_nodes->accesstype == RELATION_ACCESS_READ ||
+                       exec_nodes->accesstype == RELATION_ACCESS_READ_FQS))
                {
                        List *tmp_list = exec_nodes->nodeList;
                        exec_nodes->nodeList = GetPreferredReplicationNode(exec_nodes->nodeList);
@@ -450,7 +451,7 @@ pgxc_FQS_get_relation_nodes(RangeTblEntry *rte, Index varno, Query *query)
                        if (for_update)
                                rel_access = RELATION_ACCESS_READ_FOR_UPDATE;
                        else
-                               rel_access = RELATION_ACCESS_READ;
+                               rel_access = RELATION_ACCESS_READ_FQS;
                        break;
 
                case CMD_UPDATE:
@@ -1560,7 +1561,8 @@ pgxc_merge_exec_nodes(ExecNodes *en1, ExecNodes *en2)
        if (en1->primarynodelist || en2->primarynodelist ||
                en1->en_expr || en2->en_expr ||
                OidIsValid(en1->en_relid) || OidIsValid(en2->en_relid) ||
-               en1->accesstype != RELATION_ACCESS_READ || en2->accesstype != RELATION_ACCESS_READ)
+               (en1->accesstype != RELATION_ACCESS_READ && en1->accesstype != RELATION_ACCESS_READ_FQS) ||
+               (en2->accesstype != RELATION_ACCESS_READ && en2->accesstype != RELATION_ACCESS_READ_FQS))
                return NULL;
 
        if (IsExecNodesReplicated(en1) &&

diff --git a/src/backend/pgxc/locator/locator.c b/src/backend/pgxc/locator/locator.c
index e9c6cb82730bd5c2e8aa71e5b8c77d8a07620531..2ee9e272957eeca9141b48cb038cc89bcc0d1103 100644 (file)
--- a/src/backend/pgxc/locator/locator.c
+++ b/src/backend/pgxc/locator/locator.c
@@ -203,7 +203,8 @@ GetPreferredReplicationNode(List *relNodes)
                        break;
        }
        if (nodeid < 0)
-               return list_make1_int(linitial_int(relNodes));
+               return list_make1_int(list_nth_int(relNodes,
+                                       ((unsigned int) random()) % list_length(relNodes)));
 
        return list_make1_int(nodeid);
 }
@@ -1005,7 +1006,8 @@ createLocator(char locatorType, RelationAccessType accessType,
        {
                case LOCATOR_TYPE_REPLICATED:
                        if (accessType == RELATION_ACCESS_INSERT ||
-                                       accessType == RELATION_ACCESS_UPDATE)
+                                       accessType == RELATION_ACCESS_UPDATE ||
+                                       accessType == RELATION_ACCESS_READ_FQS)
                        {
                                locator->locatefunc = locate_static;
                                if (nodeMap == NULL)

diff --git a/src/include/pgxc/locator.h b/src/include/pgxc/locator.h
index a5501ce0d4af229752d8a5bae2d2f4cb7219ef5c..dd4892da02e51b0ce11020999fe9babbc65ff097 100644 (file)
--- a/src/include/pgxc/locator.h
+++ b/src/include/pgxc/locator.h
@@ -53,6 +53,7 @@ typedef int PartAttrNumber;
 typedef enum
 {
        RELATION_ACCESS_READ,                           /* SELECT */
+       RELATION_ACCESS_READ_FQS,                               /* SELECT for FQS */
        RELATION_ACCESS_READ_FOR_UPDATE,        /* SELECT FOR UPDATE */
        RELATION_ACCESS_UPDATE,                         /* UPDATE OR DELETE */
        RELATION_ACCESS_INSERT                          /* INSERT */