Fix handling of Immediate allocation request when user specifies processor

count (*not* node count) and exceeds the partition's node limit for a job.

src/slurmctld/node_mgr.c

@@ -1253,7 +1253,9 @@ void ping_nodes (void)
 						slurmctld_conf.last_update;
 
 		/* Request a node registration if its state is UNKNOWN or DOWN 
-		 * and periodically otherwise (about every 17th ping) */
+		 * and periodically otherwise (about every 17th ping, this 
+		 * mechanism avoids an additional timer or counter and gets 
+		 * updated configuration information once in a while) */
 		if ((base_state == NODE_STATE_UNKNOWN) || 
 		    (base_state == NODE_STATE_DOWN   ) || force_reg) {
 			debug3 ("attempt to register %s now", 

src/slurmctld/node_scheduler.c

@@ -631,9 +631,7 @@ _pick_best_nodes(struct node_set *node_set_ptr, int node_set_size,
 							*req_bitmap, min_nodes,
 							max_nodes, req_cpus, 
 							contiguous);
-			if ((pick_code == SLURM_SUCCESS) &&
-			    ((node_lim == INFINITE) ||
-			     (bit_set_count(avail_bitmap) <= node_lim)))
+			if (pick_code == SLURM_SUCCESS)
 				runable = true;
 		}
 		FREE_NULL_BITMAP(avail_bitmap);