Add a loadbalancing feature to the round-robin mapper - more to be sent to devel list

Fix a potential problem with RM-provided nodenames not matching returns from gethostname - ensure that the HNP's nodename gets DNS-resolved when comparing against RM-provided hostnames. Note that this may be an issue for RM-based clusters that don't have local DNS resolution, but hopefully that is more indicative of a poorly configured system.

This commit was SVN r18252.

orte/mca/ras/base/ras_base_node.c

@@ -23,6 +23,7 @@
 
 #include "opal/util/output.h"
 #include "opal/util/argv.h"
+#include "opal/util/if.h"
 
 #include "orte/mca/errmgr/errmgr.h"
 #include "orte/util/name_fns.h"
@@ -111,7 +112,7 @@ int orte_ras_base_node_insert(opal_list_t* nodes, orte_job_t *jdata)
          * first position since it is the first one entered. We need to check to see
          * if this node is the same as the HNP's node so we don't double-enter it
          */
-        if (0 == strcmp(node->name, hnp_node->name)) {
+        if (0 == strcmp(node->name, hnp_node->name) || opal_ifislocal(node->name)) {
             OPAL_OUTPUT_VERBOSE((5, orte_ras_base.ras_output,
                                  "%s ras:base:node_insert updating HNP info to %ld slots",
                                  ORTE_NAME_PRINT(ORTE_PROC_MY_NAME),
@@ -124,6 +125,9 @@ int orte_ras_base_node_insert(opal_list_t* nodes, orte_job_t *jdata)
             hnp_node->slots_alloc = node->slots_alloc;
             hnp_node->slots_max = node->slots_max;
             hnp_node->launch_id = node->launch_id;
+            /* use the RM's name for the node */
+            free(hnp_node->name);
+            hnp_node->name = strdup(node->name);
             /* set the node to available for use */
             hnp_node->allocate = true;
             /* update the total slots in the job */

orte/mca/rmaps/base/base.h

@@ -57,10 +57,12 @@ typedef struct {
     bool pernode;
     /** number of ppn for n_per_node mode */
     int npernode;
-    /* do we not allow use of the localhost */
+    /* do not allow use of the localhost */
     bool no_use_local;
     /* display the map after it is computed */
     bool display_map;
+    /* balance load across nodes */
+    bool loadbalance;
 } orte_rmaps_base_t;
 
 /**

orte/mca/rmaps/base/rmaps_base_open.c

@@ -125,6 +125,16 @@ int orte_rmaps_base_open(void)
         orte_rmaps_base.oversubscribe = true;
     }
     
+    /* Do we want to loadbalance the job */
+    param = mca_base_param_reg_int_name("rmaps", "base_loadbalance",
+                                        "Balance total number of procs across all allocated nodes",
+                                        false, false, (int)false, &value);
+    orte_rmaps_base.loadbalance = OPAL_INT_TO_BOOL(value);
+    /* if we are doing npernode or pernode, then we cannot loadbalance */
+    if (orte_rmaps_base.pernode) {
+        orte_rmaps_base.loadbalance = false;
+    }
+    
     /* should we display the map after determining it? */
     mca_base_param_reg_int_name("rmaps", "base_display_map",
                                 "Whether to display the process map after it is computed",

orte/mca/rmaps/base/rmaps_base_support_fns.c

@@ -88,8 +88,6 @@ int orte_rmaps_base_get_target_nodes(opal_list_t *allocated_nodes, orte_std_cntr
             ORTE_ERROR_LOG(rc);
             return rc;
         }
-    }
-    
         /** check that anything is here */
         if (0 == opal_list_get_size(allocated_nodes)) {
             opal_show_help("help-orte-rmaps-base.txt",
@@ -97,9 +95,10 @@ int orte_rmaps_base_get_target_nodes(opal_list_t *allocated_nodes, orte_std_cntr
                            true);
             return ORTE_ERR_SILENT;
         }
+    }
     
     /* did the app_context contain a hostfile? */
-    if (NULL != app->hostfile) {
+    if (NULL != app && NULL != app->hostfile) {
         /* yes - filter the node list through the file, removing
          * any nodes not found in the file
          */
@@ -108,28 +107,28 @@ int orte_rmaps_base_get_target_nodes(opal_list_t *allocated_nodes, orte_std_cntr
             ORTE_ERROR_LOG(rc);
             return rc;
         }
-    }
-    
         /** check that anything is here */
         if (0 == opal_list_get_size(allocated_nodes)) {
             opal_show_help("help-orte-rmaps-base.txt", "orte-rmaps-base:no-mapped-node",
                            true, app->app, app->hostfile);
             return ORTE_ERR_SILENT;
         }
+    }
     
    /* now filter the list through any -host specification */
+    if (NULL != app) {
         if (ORTE_SUCCESS != (rc = orte_util_filter_dash_host_nodes(allocated_nodes,
                                                                    app->dash_host))) {
             ORTE_ERROR_LOG(rc);
             return rc;
         }
-    
         /** check that anything is left! */
         if (0 == opal_list_get_size(allocated_nodes)) {
             opal_show_help("help-orte-rmaps-base.txt", "orte-rmaps-base:no-mapped-node",
                            true, app->app, "");
             return ORTE_ERR_SILENT;
         }
+    }
     
     /* If the "no local" option was set, then remove the local node
         from the list */

orte/mca/rmaps/round_robin/rmaps_rr.c

@@ -46,6 +46,7 @@
  * Local variable
  */
 static opal_list_item_t *cur_node_item = NULL;
+static int ppn = 0;
 
 /*
  * Create a default mapping for the application, scheduling round
@@ -228,10 +229,12 @@ static int map_app_by_slot(
             /* Update the number of procs allocated */
             ++num_alloc;
 
-            /** if all the procs have been mapped OR we have fully used up this node, then
-             * break from the loop
+            /** if all the procs have been mapped OR we have fully used up this node
+             * OR we are at our ppn and loadbalancing, then break from the loop 
              */
-            if(num_alloc == app->num_procs || ORTE_ERR_NODE_FULLY_USED == rc) {
+            if (num_alloc == app->num_procs ||
+                ORTE_ERR_NODE_FULLY_USED == rc ||
+                (orte_rmaps_base.loadbalance && i == ppn)) {
                 break;
             }
         }
@@ -241,7 +244,9 @@ static int map_app_by_slot(
          * node is NOT max'd out
          *
          */
-        if (i < (num_slots_to_take-1) && ORTE_ERR_NODE_FULLY_USED != rc) {
+        if (i < (num_slots_to_take-1) &&
+            ORTE_ERR_NODE_FULLY_USED != rc &&
+            i != ppn) {
             continue;
         }
         cur_node_item = next;
@@ -261,7 +266,7 @@ static int orte_rmaps_rr_map(orte_job_t *jdata)
     orte_std_cntr_t i;
     opal_list_t node_list;
     opal_list_item_t *item;
-    orte_node_t *node;
+    orte_node_t *node, **nodes;
     orte_vpid_t vpid_start;
     orte_std_cntr_t num_nodes, num_slots;
     int rc;
@@ -276,6 +281,39 @@ static int orte_rmaps_rr_map(orte_job_t *jdata)
     /* start at the beginning... */
     vpid_start = 0;
 
+    /* if loadbalancing is requested, then we need to compute
+     * the #procs/node - note that this cannot be done
+     * if we are doing pernode or if #procs was not given
+     */
+    if (orte_rmaps_base.loadbalance && !map->pernode) {
+        /* compute total #procs */
+        for(i=0; i < jdata->num_apps; i++) {
+            app = apps[i];
+            if (0 == app->num_procs) {
+                /* can't do it - just move on */
+                opal_show_help("help-orte-rmaps-rr.txt",
+                               "orte-rmaps-rr:loadbalance-and-zero-np",
+                               true);
+                rc = ORTE_ERR_SILENT;
+                goto error;
+            }
+            ppn += app->num_procs;
+        }
+        /* get the total avail nodes */
+        nodes = (orte_node_t**)orte_node_pool->addr;
+        num_nodes=0;
+        for (i=0; i < orte_node_pool->size; i++) {
+            if (NULL == nodes[i]) {
+                break;  /* nodes are left aligned, so stop when we hit a null */
+            } 
+            if (nodes[i]->allocate) {
+                num_nodes++;
+            }
+        }
+        /* compute the balance */
+        ppn = ppn / num_nodes;
+    }
+
     /* cycle through the app_contexts, mapping them sequentially */
     for(i=0; i < jdata->num_apps; i++) {
         app = apps[i];

orte/tools/orterun/orterun.c

@@ -215,6 +215,9 @@ static opal_cmd_line_init_t cmd_line_init[] = {
     { "rmaps", "base", "no_oversubscribe", '\0', "nooversubscribe", "nooversubscribe", 0,
       NULL, OPAL_CMD_LINE_TYPE_BOOL,
       "Nodes are not to be oversubscribed, even if the system supports such operation"},
+    { "rmaps", "base", "loadbalance", '\0', "loadbalance", "loadbalance", 0,
+      NULL, OPAL_CMD_LINE_TYPE_BOOL,
+      "Balance total number of procs across all allocated nodes"},
     { "rmaps", "base", "display_map", '\0', "display-map", "display-map", 0,
       NULL, OPAL_CMD_LINE_TYPE_BOOL,
       "Display the process map just before launch"},