svn merge -r17864:17869 https://eris.llnl.gov/svn/slurm/branches/slurm-2.0

NEWS

@@ -90,8 +90,12 @@ documents those changes that are of interest to users and admins.
     failed.
  -- BLUEGENE - Fix, for if you are setting an subbp into an error mode 
     where the subbp stated isn't the first ionode in a nodecard.
- -- fix for backfill to not core when checking shared nodes.
- -- fix for scontrol to not core when hitting just return in interactive mode.
+ -- Fix for backfill to not core when checking shared nodes.
+ -- Fix for scontrol to not core when hitting just return in interactive mode.
+ -- Improve sched/backfill logic with respect to shared nodes (multiple jobs
+    per node).
+ -- In sched/wiki (Maui interface) add job info fields QOS, RCLASS, DMEM and
+    TASKSPERNODE.
 
 * Changes in SLURM 2.0.1
 ========================

doc/html/team.shtml

@@ -50,6 +50,7 @@ Linux NetworX and many other contributorss.
 <li>Donald Lipari (LLNL)</li>
 <li>Steven McDougall (SiCortex)</li>
 <li>Donna Mecozzi (LLNL)</li>
+<li>Bj&oslash;rn-Helge Mevik (University of Oslo, Norway)</li>
 <li>Chris Morrone (LLNL)</li>
 <li>Pere Munt (Barcelona Supercomputer Center, Spain)</li>
 <li>Michal Novotny (Masaryk University, Czech Republic)</li> 
@@ -76,6 +77,6 @@ Networking, Italy)</li>
 <li>Anne-Marie Wunderlin (Bull)</li>
 </ul>
 
-<p style="text-align:center;">Last modified 27 May 2009</p>
+<p style="text-align:center;">Last modified 16 June 2009</p>
 
 <!--#include virtual="footer.txt"-->

doc/man/man1/salloc.1

@@ -87,7 +87,7 @@ For example:
    \-\-begin=16:00
    \-\-begin=now+1hour
    \-\-begin=now+60           (seconds by default)
-   \-\-begin=2010-01-20T12:34:67
+   \-\-begin=2010-01-20T12:34:00
 .fi
 
 .TP

doc/man/man1/sbatch.1

@@ -86,7 +86,7 @@ For example:
    \-\-begin=16:00
    \-\-begin=now+1hour
    \-\-begin=now+60           (seconds by default)
-   \-\-begin=2010-01-20T12:34:67
+   \-\-begin=2010-01-20T12:34:00
 .fi
 
 .TP

doc/man/man1/srun.1

@@ -74,7 +74,7 @@ For example:
    \-\-begin=16:00
    \-\-begin=now+1hour
    \-\-begin=now+60           (seconds by default)
-   \-\-begin=2010-01-20T12:34:67
+   \-\-begin=2010-01-20T12:34:00
 .fi
 
 .TP

src/plugins/sched/backfill/backfill.c

@@ -248,6 +248,7 @@ static int  _try_sched(struct job_record *job_ptr, bitstr_t **avail_bitmap,
 		/* Try to schedule the job. First on dedicated nodes
 		 * then on shared nodes (if so configured). */
 		uint16_t orig_shared;
+		time_t now = time(NULL);
 		orig_shared = job_ptr->details->shared;
 		job_ptr->details->shared = 0;
 		tmp_bitmap = bit_copy(*avail_bitmap);
@@ -255,7 +256,8 @@ static int  _try_sched(struct job_record *job_ptr, bitstr_t **avail_bitmap,
 				       max_nodes, req_nodes,
 				       SELECT_MODE_WILL_RUN);
 		job_ptr->details->shared = orig_shared;
-		if ((rc != SLURM_SUCCESS) && (orig_shared != 0)) {
+		if (((rc != SLURM_SUCCESS) || (job_ptr->start_time > now)) &&
+		    (orig_shared != 0)) {
 			FREE_NULL_BITMAP(*avail_bitmap);
 			*avail_bitmap= tmp_bitmap;
 			rc = select_g_job_test(job_ptr, *avail_bitmap, 

src/plugins/sched/wiki/get_jobs.c

@@ -17,7 +17,7 @@
  *  any later version.
  *
  *  In addition, as a special exception, the copyright holders give permission 
- *  to link the code of portions of this program with the OpenSSL library under 
+ *  to link the code of portions of this program with the OpenSSL library under
  *  certain conditions as described in each individual source file, and 
  *  distribute linked combinations including the two. You must obey the GNU 
  *  General Public License in all respects for all of the code used other than 
@@ -51,6 +51,7 @@
 static char *	_dump_all_jobs(int *job_cnt, time_t update_time);
 static char *	_dump_job(struct job_record *job_ptr, time_t update_time);
 static uint16_t _get_job_cpus_per_task(struct job_record *job_ptr);
+static uint16_t _get_job_tasks_per_node(struct job_record *job_ptr);
 static uint32_t	_get_job_end_time(struct job_record *job_ptr);
 static char *	_get_job_features(struct job_record *job_ptr);
 static uint32_t	_get_job_min_disk(struct job_record *job_ptr);
@@ -83,16 +84,19 @@ static char *	_task_list(struct job_record *job_ptr);
  *	WCLIMIT=<secs>;			wall clock time limit, seconds
  *	TASKS=<cpus>;			CPUs required
  *	[NODES=<nodes>;]		count of nodes required
+ *	[TASKSPERNODE=<cnt>;]		tasks required per node
  *	DPROCS=<cpus_per_task>;		count of CPUs required per task
  *	QUEUETIME=<uts>;		submission time
  *	STARTTIME=<uts>;		time execution started
  *	PARTITIONMASK=<partition>;	partition name
+ *	[DMEM=<mbytes>;]		MB of memory required per cpu
  *	RMEM=<MB>;			MB of memory required
  *	RDISK=<MB>;			MB of disk space required
  *	[COMPLETETIME=<uts>;]		termination time
  *	[SUSPENDTIME=<secs>;]		seconds that job has been suspended
- *	[QOS=<quality_of_service>];	quality of service
- *	[ACCOUNT=<bank_account>];	bank account name
+ *	[ACCOUNT=<bank_account>;]	bank account name
+ *	[QOS=<quality_of_service>;]	quality of service
+ *	[RCLASS=<resource_class>;]	resource class
  *	[COMMENT=<whatever>;]		job dependency or account number
  *	UNAME=<user_name>;		user name
  *	GNAME=<group_name>;		group name
@@ -205,7 +209,7 @@ static char *	_dump_job(struct job_record *job_ptr, time_t update_time)
 {
 	char tmp[16384], *buf = NULL;
 	char *uname, *gname;
-	uint32_t end_time, suspend_time;
+	uint32_t end_time, suspend_time, min_mem;
 
 	if (!job_ptr)
 		return NULL;
@@ -266,10 +270,18 @@ static char *	_dump_job(struct job_record *job_ptr, time_t update_time)
 	xstrcat(buf, tmp);
 
 	if (!IS_JOB_FINISHED(job_ptr)) {
+	        uint16_t tpn;
 		snprintf(tmp, sizeof(tmp),
 			"NODES=%u;",
 			_get_job_min_nodes(job_ptr));
 		xstrcat(buf, tmp);
+		tpn = _get_job_tasks_per_node(job_ptr);
+		if (tpn > 0) {
+			snprintf(tmp, sizeof(tmp),
+				 "TASKPERNODE=%u;",
+				 tpn);
+			xstrcat(buf, tmp);
+		}
 	}
 
 	snprintf(tmp, sizeof(tmp),
@@ -284,6 +296,13 @@ static char *	_dump_job(struct job_record *job_ptr, time_t update_time)
 		job_ptr->partition);
 	xstrcat(buf, tmp);
 
+	min_mem = _get_job_min_mem(job_ptr);
+	if (min_mem & MEM_PER_CPU) {
+		snprintf(tmp, sizeof(tmp),
+			"DMEM=%u;", min_mem & (~MEM_PER_CPU));
+		xstrcat(buf, tmp);
+	}
+
 	snprintf(tmp, sizeof(tmp),
 		"RMEM=%u;RDISK=%u;",
 		_get_job_min_mem(job_ptr),
@@ -306,7 +325,7 @@ static char *	_dump_job(struct job_record *job_ptr, time_t update_time)
 
 	if (job_ptr->account) {
 		/* allow QOS spec in form "qos-name" */
-		if (!strncmp(job_ptr->account,"qos-",4)) {
+		if (!strncmp(job_ptr->account, "qos-", 4)) {
 			snprintf(tmp, sizeof(tmp),
 				 "QOS=%s;", job_ptr->account + 4);
 		} else {
@@ -317,9 +336,33 @@ static char *	_dump_job(struct job_record *job_ptr, time_t update_time)
 	}
 
 	if (job_ptr->comment && job_ptr->comment[0]) {
-		snprintf(tmp,sizeof(tmp),
-			"COMMENT=%s;", job_ptr->comment);
-		xstrcat(buf,tmp);
+		/* Parse comment for class/qos spec */
+		char *copy;
+		char *cred, *value;
+		copy = xstrdup(job_ptr->comment);
+		cred = strtok(copy, ",");
+		while (cred != NULL) {
+			if (!strncmp(cred, "qos:", 4)) {
+				value = &cred[4];
+				if (value[0] != '\0') {
+					snprintf(tmp, sizeof(tmp),
+						 "QOS=%s;", value);
+					xstrcat(buf, tmp);
+				}
+			} else if (!strncmp(cred, "class:", 6)) {
+				value = &cred[6];
+				if (value[0] != '\0') {
+					snprintf(tmp, sizeof(tmp),
+						"RCLASS=%s;", value);
+					xstrcat(buf, tmp);
+				}
+			}
+			cred = strtok(NULL, ",");
+		}
+		xfree(copy);
+		snprintf(tmp, sizeof(tmp),
+			 "COMMENT=%s;", job_ptr->comment);
+		xstrcat(buf, tmp);
 	}
 
 	if (job_ptr->details &&
@@ -346,6 +389,16 @@ static uint16_t _get_job_cpus_per_task(struct job_record *job_ptr)
 	return cpus_per_task;
 }
 
+
+static uint16_t _get_job_tasks_per_node(struct job_record *job_ptr)
+{
+	uint16_t tasks_per_node = 0;
+
+	if (job_ptr->details && job_ptr->details->ntasks_per_node)
+		tasks_per_node = job_ptr->details->ntasks_per_node;
+	return tasks_per_node;
+}
+
 static uint32_t _get_job_min_mem(struct job_record *job_ptr)
 {
 	if (job_ptr->details)