FEATURE: Added RMCFG[] ASYNCJOBSTART=TRUE for asynchronous job starts in pbs.

Problem Ref:	MAUI + TORQUE JOB START RATE
Organization:	NA
Contact:	NA



git-svn-id: svn://opensvn.adaptivecomputing.com/maui/trunk@115 3f5042e3-fb1d-0410-be18-d6ca2573e517

CHANGELOG

@@ -1,4 +1,5 @@
 Maui 3.2.6p21
+  - Added RMCFG[] ASYNCJOBSTART=TRUE for asynchronous job starts in pbs. (Thanks to Bas van der Vlies and the community)
   - Added StartTime and CompletionTime to Gold Charge.
   - Fixed backfill issue with SINGLEUSER NODEACCESSPOLICY. (Thanks goes to Roy Dragseth)
   - N->{A|C}Res.Swap is overcommitted with N->CRes.Swap instead of N->CRes.Mem. (Thanks goes to Roy Dragseth)

include/moab.h

@@ -1013,6 +1013,7 @@ enum MLimitAttrType {
 enum MRMAttrType {
   mrmaNONE = 0,
   mrmaAuthType,
+  mrmaASyncJobStart, /* asynchronous job start */
   mrmaConfigFile,
   mrmaCSAlgo,
   mrmaCSKey,

include/msched.h

@@ -1136,6 +1136,8 @@ typedef struct {
   void *S;  /* resource manager specific data */
 
   void *xd;
+
+  mbool_t ASyncJobStart; /* asynchronous job start */
   } mrm_t;
 
 enum MClassStateEnum {

src/moab/MConst.c

@@ -441,6 +441,7 @@ const char *MFSAttr[] = {
 const char *MRMAttr[] = {
   NONE,
   "AUTHTYPE",
+  "ASYNCJOBSTART",
   "CONFIGFILE",
   "CSALGO",
   "CSKEY",

src/moab/MPBSI.c

@@ -1796,41 +1796,6 @@ int MPBSJobStart(
 
       return(FAILURE);
       }
-
-    if (MPBSJobModify(
-          J,
-          R,
-          ATTR_l,
-          (char *)(R->Version >= 710 ? "select" : "neednodes"),
-          HostList,
-          NULL,
-          NULL) == FAILURE)
-      {
-      DBG(0,fPBS) DPrint("ERROR:    cannot set hostlist for job '%s'\n",
-        J->Name);
-
-      if (R->FailIteration != MSched.Iteration)
-        {
-        R->FailIteration = MSched.Iteration;
-        R->FailCount     = 0;
-        }
-
-      R->FailCount++;
-
-      if (Msg != NULL)
-        strcpy(Msg,"job cannot be started - cannot set hostlist");
-
-      if (SC != NULL)
-        *SC = mscRemoteFailure;
-
-      return(FAILURE);
-      }
-    else
-      {
-      DBG(7,fPBS) DPrint("INFO:     hostlist for job '%s' set to '%s'\n",
-        J->Name,
-        HostList);
-      }
     }
   else
     {
@@ -1909,7 +1874,16 @@ int MPBSJobStart(
 
   MJobGetName(J,NULL,R,tmpJobName,sizeof(tmpJobName),mjnRMName);       
 
-  rc = pbs_runjob(R->U.PBS.ServerSD,tmpJobName,MasterHost,NULL);
+  if (R->ASyncJobStart == FALSE )
+    {
+    DBG(7,fPBS) DPrint("INFO:     use pbs_runjob\n");
+    rc = pbs_runjob(R->U.PBS.ServerSD,tmpJobName,HostList,NULL);
+    }
+  else
+    {
+    DBG(7,fPBS) DPrint("INFO:     use pbs_asyrun\n");
+    rc = pbs_asyrunjob(R->U.PBS.ServerSD,tmpJobName,HostList,NULL);
+    }
 
   if (rc != 0)
     {
@@ -1933,28 +1907,6 @@ int MPBSJobStart(
     JobStartFailed = TRUE;
     }
 
-  if (J->NeedNodes != NULL)
-    {
-    if (MPBSJobModify(
-          J,
-          R,
-          ATTR_l,
-          (char *)(R->Version >= 710 ? "select" : "neednodes"),
-          J->NeedNodes,
-          NULL,
-          NULL) == FAILURE)
-      {
-      DBG(7,fPBS) DPrint("WARNING:  cannot reset hostlist for job '%s')\n",
-        J->Name);
-      }
-    else
-      {
-      DBG(7,fPBS) DPrint("INFO:     hostlist for job '%s' set to '%s'\n",
-        J->Name,
-        J->NeedNodes);
-      }
-    }
-
   if (JobStartFailed == TRUE)
     {
     /* job could not be started */

src/moab/MRM.c

@@ -1771,7 +1771,7 @@ int MRMLoadConfig(
 
         MRMSetDefaults(R);
 
-	MOLoadPvtConfig((void **)R,mxoRM,NULL,NULL,NULL);
+        MOLoadPvtConfig((void **)R,mxoRM,NULL,NULL,NULL);
         }
  
       /* load RM specific attributes */
@@ -2032,6 +2032,8 @@ int MRMSetDefaults(
 
     R->JobCounter = 0;
     
+    R->ASyncJobStart = FALSE;
+
     return(SUCCESS);
     }
  
@@ -2242,7 +2244,7 @@ int MRMProcessConfig(
           ptr,
           (const char **)MRMAttr,
           &aindex,
-	  NULL,
+          NULL,
           TRUE,
           NULL,
           ValLine,
@@ -2266,6 +2268,12 @@ int MRMProcessConfig(
 
         break;
 
+      case mrmaASyncJobStart:
+
+        R->ASyncJobStart = MUBoolFromString(ValLine,FALSE);
+        
+        break;
+
       case mrmaConfigFile:
 
         MUStrCpy(R->U.LL.ConfigFile,ValLine,sizeof(R->U.LL.ConfigFile));