Added logic to return scheduled nodes to Maui scheduler (David

Jackson, Cluster Resources)

NEWS

@@ -14,6 +14,8 @@ documents those changes that are of interest to users and admins.
     add string length checking.
  -- Modify orphan batch job time calculation for BGL to account for 
     slowness when booting many bglblocks at the same time.
+ -- Added logic to return scheduled nodes to Maui scheduler (David
+    Jackson, Cluster Resources)
 
 * Changes in SLURM 0.6.0-pre4
 =============================

src/plugins/sched/maui/wiki/wiki_canceljob.cpp

@@ -73,5 +73,5 @@ wiki_canceljob_t::action( void )
 	u_int32_t id = (u_int32_t) atol( m_jobid );
 	verbose( "Wiki canceling job %s", m_jobid );
 	return new wiki_status_t( ( sched_cancel_job( id )
-				    == SLURM_SUCCESS ) ? 0 : 1 );
+				    == SLURM_SUCCESS ) ? 0 : -1 );
 }

src/plugins/sched/maui/wiki/wiki_request.cpp

@@ -521,6 +521,7 @@ wiki_getjobs_t::wiki_getjobs_t( char *data, size_t len ) :
 		// JOB_FIELD_END_TIME,  -- this confuses Maui
 		JOB_FIELD_USER_ID,
 		JOB_FIELD_GROUP_ID,
+		JOB_FIELD_ALLOC_NODES,
 		JOB_FIELD_REQ_NODES,
 		JOB_FIELD_PARTITION,
 		JOB_FIELD_MIN_NODES,
@@ -546,6 +547,7 @@ wiki_getjobs_t::wiki_getjobs_t( char *data, size_t len ) :
 		{ "UNAME",		JOB_FIELD_USER_ID },
 		{ "GNAME",		JOB_FIELD_GROUP_ID },
 		{ "HOSTLIST",		JOB_FIELD_REQ_NODES },
+		{ "TASKLIST",           JOB_FIELD_ALLOC_NODES },
 		{ "PARTITIONMASK",     	JOB_FIELD_PARTITION },
 		{ "NODES",		JOB_FIELD_MIN_NODES },
 		{ "RMEM",		JOB_FIELD_MIN_MEMORY },
@@ -574,7 +576,7 @@ wiki_getjobs_t::map_enum( char * const field,
 	static struct string_map job_state_map[] = {
 		{ JOB_STATE_LABEL_PENDING,	"Idle" },
 		{ JOB_STATE_LABEL_RUNNING,    	"Running" },
-		{ JOB_STATE_LABEL_COMPLETE,	"Complete" },
+		{ JOB_STATE_LABEL_COMPLETE,	"Completed" },
 		{ JOB_STATE_LABEL_FAILED,    	"Removed" },
 		{ JOB_STATE_LABEL_TIMEOUT,	"Removed" },
 		{ JOB_STATE_LABEL_NODE_FAIL,	"Removed" },

src/plugins/sched/maui/wiki/wiki_startjob.cpp

@@ -68,6 +68,7 @@ message_t *
 wiki_startjob_t::action( void )
 {
 	u_int32_t id = (u_int32_t) atol( m_jobid );
+	int rc;
 
 	// *
 	// If Maui has specified a node list to run on, change the
@@ -95,6 +96,15 @@ wiki_startjob_t::action( void )
 	}
 	
 	verbose( "Wiki starting job %s", m_jobid );	
-	return new wiki_status_t( (sched_start_job( id, (u_int32_t) 1 )
-				   == SLURM_SUCCESS ) ? : 0 -1 );
+
+	rc = sched_start_job( id, (u_int32_t) 1 );
+
+	if (rc == SLURM_SUCCESS)
+	{
+		return new wiki_status_t( 0, "SUCCESS: job %s started successfully");
+	}
+	else
+	{
+		return new wiki_status_t( -1, "ERROR: job %s failed to start");
+	}
 }

src/plugins/sched/maui/wiki/wiki_status.cpp

@@ -52,6 +52,8 @@ wiki_status_t::wiki_status_t( int status, char * const msg ) :
 		m_str += " RESPONSE=";
 		m_str += msg;
 	}
+
+	debug3( "Wiki plugin status = \"%s\"", m_str.s() );
 }
 
 // **************************************************************

src/slurmctld/sched_plugin.h

@@ -289,6 +289,7 @@ extern sched_obj_list_t sched_get_job_list( void );
 #define JOB_FIELD_MIN_DISK 		"job.min_disk"
 #define JOB_FIELD_MIN_MEMORY	       	"job.min_mem"
 #define JOB_FIELD_REQ_NODES		"job.req_nodes"
+#define JOB_FIELD_ALLOC_NODES		"job.alloc_nodes"
 #define JOB_FIELD_MIN_NODES		"job.min_nodes"
 
 #define JOB_STATE_LABEL_PENDING		"PENDING"

src/slurmctld/sched_upcalls.c

@@ -116,6 +116,7 @@ static void * sched_get_job_end_time(	sched_obj_list_t, int32_t, char * );
 static void * sched_get_job_user_id(	sched_obj_list_t, int32_t, char * );
 static void * sched_get_job_group_name(	sched_obj_list_t, int32_t, char * );
 static void * sched_get_job_req_nodes(	sched_obj_list_t, int32_t, char * );
+static void * sched_get_job_alloc_nodes( sched_obj_list_t, int32_t, char * );
 static void * sched_get_job_min_nodes(	sched_obj_list_t, int32_t, char * );
 static void * sched_get_job_partition(	sched_obj_list_t, int32_t, char * );
 static void * sched_get_job_min_disk( sched_obj_list_t, int32_t, char * );
@@ -225,6 +226,12 @@ sched_get_accessor( char *field )
 		{ JOB_FIELD_END_TIME,		sched_get_job_end_time },
 		{ JOB_FIELD_USER_ID,		sched_get_job_user_id },
 		{ JOB_FIELD_GROUP_ID,		sched_get_job_group_name },
+		/*  { JOB_FIELD_ALLOC_NODES,	sched_get_job_alloc_nodes },
+		 * Wiki specifies the nodes to be allocated in the requested node
+		 * field, so that is where we are getting the allocated node 
+		 * information from for now.
+		 */
+		{ JOB_FIELD_ALLOC_NODES,	sched_get_job_req_nodes },
 		{ JOB_FIELD_REQ_NODES,		sched_get_job_req_nodes },
 		{ JOB_FIELD_MIN_NODES,		sched_get_job_min_nodes },
 		{ JOB_FIELD_PARTITION,		sched_get_job_partition },
@@ -745,6 +752,36 @@ sched_get_job_req_nodes( sched_obj_list_t job_data,
 	return "";
 }
 
+
+/* ************************************************************************ */
+/*  TAG(                       sched_get_job_alloc_nodes                 )  */
+/* ************************************************************************ */
+static void *
+sched_get_job_alloc_nodes( sched_obj_list_t job_data,
+                         int32_t idx,
+                         char *type )
+{
+        void *cache;
+        char *nodes;
+
+        if ( type ) *type = 'S';
+        nodes = ( (struct job_record *)job_data->data )[ idx ].nodes;
+
+        if ( nodes ) {
+                if ( ( cache = sched_obj_cache_entry_find( job_data,
+                                                           idx,
+                                                           "alloc_nodes" ) ) != NULL ) {
+                        return cache;
+                }
+                cache = expand_hostlist( nodes );
+                if ( ! cache )
+                        return nodes;
+                sched_obj_cache_entry_add( job_data, idx, "alloc_nodes", cache );
+                return cache;
+        }
+        return "";
+}
+
 /* ************************************************************************ */
 /*  TAG(                       sched_get_job_min_nodes                   )  */
 /* ************************************************************************ */